PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProlecc&Prele|RC/DateDue.indd

 

BIOMOLECULAR ENGINEERING OF SIRNA
THERAPEUTICS

By

JOSEPH A. GREDELL

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Chemical Engineering

2009

ABSTRACT
BIOMOLECULAR ENGINEERING OF SIRNA THERAPEUTICS
By
JOSEPH A. GREDELL

RNA interference (RN Ai) provides a powerful means for regulating gene
expression. Although it is a relatively recent discovery, it has already proven useful in
identification of gene/protein fimctions and is in the process of being utilized for disease
treatment in humans. Continued use of this pathway for a variety of applications would
benefit from a more thorough understanding of the role of the initiator molecules of
RNAi, small interfering RNA (siRNA). As the efficacy of siRNAs that mediate RNAi is
known to vary greatly, the aim of this thesis was to study three steps of the pathway that
pose barriers to identifying and developing highly active siRNA molecules for use as
therapeutic agents.

The first of these factors investigated in this research was the effect of secondary
structure within the mRN A site targeted by siRNA. The results obtained from both
experimental and computational studies show that sites with extensive mRN A secondary
structure were less susceptible to silencing, while those containing unpaired nucleotides
at either the 5'— or 3'-end were generally more amenable. Similar observations were made
when taking into account siRN A guide strand structure. Taken together, there is a
correlation between RNA structure and silencing efficiency that ultimately can be
included in existing and fi1ture siRN A selection algorithms for the improved
identification of active siRNAs, thus reducing the number of sequences to be tested

before selecting one capable of significantly silencing the target gene.

The second aspect of RNAi explored here was the influence of siRNA sequence
and hybridization stability on recognition by the TAR RNA binding protein (TRBP).
This protein is part of the RNA induced silencing complex (RISC) responsible for
targeting and cleaving mRNAs to achieve silencing. It was found that TRBP can detect
the overall and relative stability of the two ends of the siRNA by interacting primarily
with the more stable end, and that the interaction is similarly reflected when binding to
single-stranded RNAs alone. Additionally, TRBP interactions can be altered through the
inclusion of mismatches within the siRNA sequence or DNA substitutions. These
observations suggest a role for TRBP in the mechanism of RISC formation as well as a
means to improve siRNA functionality through modification of the RNA or its sequence.

The third focus was to develop a biocompatible and biodegradable cationic
polymer capable of delivering siRNA into cells grown in vitro, with potential use later for
in viva applications. These nanoparticles (NPs) were generated with “clic ” chemistry,
allowing for a systematic study of a range of variables contributing to siRNA binding.
Strong siRN A binding required a NP containing a combination of primary and secondary
amine groups to facilitate electrostatic interactions. Inclusion of alkyl chains enhanced
binding, perhaps by causing vesicle-like formation, while polyethylene glycol (PEG)
reduced overall binding affinity, consistent with its known charge shielding effects. The
NPs were capable of delivering siRNAs to cells, but were unable to release them to
initiate RNAi. Collectively, the work presented here provides a framework for
identifying and engineering more active siRNA molecules by taking into account mRN A
target structure and TRBP binding preferences, as well as achieving efficient cellular

uptake of such engineered molecules with a novel type of polymeric nanoparticle.

C0pyright by
JOSEPH A. GREDELL

2009

To My Family

ACKNOWLEDGEMENTS

I thank the many people who, in some form or another, contributed to the work
described in this document.

First and foremost, I am grateful to my advisor, Professor Pat Walton, for his
constant support and guidance. I am certain that his mentoring has made me into a better
scientist, and person, that will serve me well in the future.

I also thank Professor Kris Chan, for closely assisting with my work over the
years, in addition to Professors Mark Worden and Richard Schwartz, for serving as
members of my committee; thank you for your critical, yet helpful, discussions, as well
as for having taught several of the courses I enrolled in as part of my Ph.D. program. I
am also thankful for the collaboration with Professor Greg Baker and Erin Vogel as we
studied the polymeric nanoparticles that, although they "work" in many regards, are still a
source of frustration.

Of course, I would be remiss if I didn't express my gratitude to the members of
the Chan and Walton labs who I worked with on a daily basis and who provided support,
critical analysis, and discussions. A special note of thanks goes out to Shireesh, Sri,
Xuerui, Hemant, Shengnan, and Mike. In addition, I am grateful for the help of
numerous undergraduate students who assisted with many of the daily experiments and
were a pleasure to work with: Greeshma, Sophie, Dan, Mark, Jorge, Melissa, and in
particular, Angela and Mike (Dittmer).

I acknowledge Michigan State University, the College of Engineering, the
Department of Chemical Engineering and Materials Science, as well as the National

Science Foundation and the National Institutes of Health for financial support.

I was fortunate to meet many other great people along the way, who helped me
with my work, but more importantly were there as fiiends. In particular, Aaron, Brian,
and Susan — thank you for listening, especially towards the end. And thank you to the
MSU Men's Volleyball Club for showing me what a college experience really can be like.

Last but not least, I thank my mom, dad, and sister. There is too much to say, and
too few words. But I would not be here today if not for your continuous love and support

from the beginning.

Thank you.

vii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................... x
LIST OF FIGURES ............................................................................................................ xi
LIST OF ABBREVIATIONS ......................................................................................... xiii
CHAPTER 1. INTRODUCTION ........................................................................................ 1
1.1 Significance ................................................................................................................... 1
1.2 Background .................................................................................................................... 2
1.3 RNA Interference mechanism ....................................................................................... 3
1.3.1 Human RNAi mechanism ....................................................................................... 8
1.3.2 dsRBPs involved in and related to RNAi ............................................................... 9
1.3.3 Non-specific effects - Immune response and off-target effects of siRNAs .......... 14
1.4 Application of RNAi ................................................................................................... 15
1.4.1 Selection of highly active siRNA sequences ........................................................ 15
1.4.2 Specific uses of RNAi .......................................................................................... 19
1.4.3 siRNA delivery ..................................................................................................... 22
1.4.4 Chemical modification of siRNA ......................................................................... 27
1.5 Approach and specific aims ......................................................................................... 29
CHAPTER 2. IMPACT OF TARGET MRNA STRUCTURE ON SIRNA SILENCING
EFFICIENCY .................................................................................................................... 32
2.1 Abstract ........................................................................................................................ 32
2.2 Introduction ................................................................................................................. 34
2.3 Results ......................................................................................................................... 36
2.3.1 siRNA selection and mRN A target structure determination ................................ 36
2.3.2 Silencing activity of siRNAs targeting the EGFP mRNA .................................... 41
2.3.3 Consideration of larger siRN A dataset — data distribution ................................... 41
2.3.4 Distribution of base-pairs within the mRN A target and the effect on silencing
activity ........................................................................................................................... 43
2.3.5 mRNA target loop size and location effects ......................................................... 46
2.3.6 siRNA guide strand structure effects .................................................................... 49
2.4 Discussion .................................................................................................................... 52
CHAPTER 3 RECOGNITION OF SIRNA BY TAR RNA BINDING PROTEIN (TRBP)
........................................................................................................................................... 61
3.1 Abstract ........................................................................................................................ 61
3.2 Introduction ................................................................................................................. 62
3.3 Results ......................................................................................................................... 64
3.3.1 TRBP recognition of small nucleic acids ............................................................. 65
3.3.2 TRBP asymmetry sensing in siRNAs ................................................................... 70
3.3.3 Altering siRNA asymmetry via mismatches ........................................................ 76
3.3.4 TRBP recognition of single-stranded small RNAs ............................................... 82
3.4 Discussion .................................................................................................................... 85

viii

CHAPTER 4 POLYMERIC NAN OPARTICLES FOR SIRNA DELIVERY .................. 91

4.1 Abstract ........................................................................................................................ 91
4.2 Introduction ................................................................................................................. 93
4.3 Results and Discussion ................................................................................................ 95
CHAPTER 5 CONCLUSIONS AND FUTURE WORK ................................................ 116
5.1 Conclusions ............................................................................................................... l 16
5.2 Future Work ............................................................................................................... 117

5.2.1 siRNA selection algorithm ................................................................................. 118

5.2.2 Further TRBP characterization ........................................................................... 120

5.2.3 Characterization of additional RNAi-related proteins ........................................ 121

5.2.4 Further nanoparticle investigations .................................................................... 123
APPENDICES ................................................................................................................. 125
Materials and Methods for Ch. 2 ..................................................................................... 125
Materials and Methods for Ch. 3 ..................................................................................... 128
Materials and Methods for Ch. 4 ..................................................................................... 137
BIBLIOGRAPHY ........................................................................................................... 141

ix

LIST OF TABLES

Table 1-1 siRNAs in clinical trials (http://clinicaltrialsgov). ........................................... 21
Table 2-1 Details for EGFP targeting siRNAs experimentally validated in this study. ....38
Table 2-2 Definition of RNA structures and siRNA activity within each group. ............. 48

Table 2-3 Statistical analysis of siRNA activity between groups for mRNA target
structures ............................................................................................................................ 50

Table 2-4 Statistical analysis of siRNA activity between groups for guide strand
structures ............................................................................................................................ 51

Table 2-5 Distribution of siRNA functionality for each target structure classification. ...53

Table 4-1 Binding strength (K) and cooperativity (n) of nanoparticles for siDNA as
determined by gel shift assay ........................................................................................... 101

Table 6-1 Nucleic acids used in Ch. 3 (listed 5' to 3'). .................................................... 130

Table 6-2 Nucleic acid duplexes used in Ch. 3 (top strand listed 5' to 3', bottom strand
listed 3' to 5') .................................................................................................................... 131

Table 6-3 Nucleic acid duplexes targeting pp-luciferase used in Ch. 3 (top strand listed 5'
to 3', bottom strand listed 3' to 5'). ................................................................................... 132

Table 6-4 Nucleic acid duplexes targeting sodl used in Ch. 3 (top strand listed 5' to 3',
bottom strand listed 3' to 5') ............................................................................................. 133

Table 6-5 Nucleic acid duplexes targeting EGFP used in Ch. 3 (top strand listed 5' to 3',
bottom strand listed 3' to 5') ............................................................................................. 134

LIST OF FIGURES

Images in this dissertation are presented in color.

Figure 1-1 Mechanism of RNAi. ......................................................................................... 4
Figure 1-2 siRNA structure. ................................................................................................ 6
Figure 1-3 Double stranded RNA binding domain structures of three dsRBPs. ............... 10
Figure 1-4 TAR RNA stem-loop secondary structure. ...................................................... 12
Figure 2-1 mRNA predicted structures for siRNAs targeting EGFP. ............................... 39
Figure 2-2 Local mRNA target structure groupings .......................................................... 40
Figure 2-3 mRNA target structure dependent gene silencing. .......................................... 42
Figure 2-4 Distribution of siRNAs. ................................................................................... 44
Figure 2-5 Influence of the number of base-pairs within the target site ............................ 45
Figure 2-6 Effect of window size on gene expression level. ............................................. 47
Figure 2-7 siRNA-mRNA interaction predictions ............................................................. 60
Figure 3-1 Characterization of recombinant TAR RNA binding protein (T RBP). ........... 66
Figure 3-2 TRBP binding preferences for nucleic acids. .................................................. 67
Figure 3-3 TRBP multimcrization on siRNA. ................................................................... 68
Figure 3-4 No MBP binding of small nucleic acids. ......................................................... 69
Figure 3-5 Nucleic acids used. .......................................................................................... 71
Figure 3-6 TRBP asymmetry sensing ................................................................................ 73
Figure 3-7 Asymmetry sensing by human cell extract. ..................................................... 75
Figure 3-8 Computational analysis of first nucleotide influence on siRNA activity. ....... 77
Figure 3-9 Statistical analysis for pair-wise comparison between first nucleotide on either
end of the siRNAs .............................................................................................................. 78
Figure 3-10 Mismatched siRNAs alter TRBP asymmetry. ............................................... 80

xi

Figure 3-11 TRBP recognition of short single-stranded RNAs. ....................................... 83

Figure 4-1 Nanoparticle synthesis details .......................................................................... 96
Figure 4.2 Visualization of nanoparticle delivery into cell culture by fluorescence

microscopy ......................................................................................................................... 98
Figure 4—3 Nanoparticle complex formation with siDNA (analog of siRNA). ............... 100

Figure 4-4 Nanoparticle functional groups added to the propargyl glycolide backbone. 1 04

Figure 4-5 Nanoparticle binding affinity relative to polymer backbone length. ............. 107
Figure 4-6 Nanoparticle binding affinity relative to PEG length. ................................... 111
Figure 4-7 Nanoparticle binding affinity relative to alkyl chain length. ....................... 112
Figure 4-8 Nanoparticle size determined by Dynamic Light Scattering (DLS) .............. 113
Figure 4-9 Nanoparticle size and shape determined by Transmission Electron Microscopy
(TEM). ............................................................................................................................. l 14
Figure 4-10 Nanoparticle toxicity .................................................................................... 115

xii

Ago

AS
asODN
ATP
BLAST
BPEI
CHO
dFXR
DLS
DNA
DOTMA
dsDNA
dsRBD
dsRBM
dsRBP
dsRNA
EGFP
eIFZa
FAM
FIT C

FPLC

LIST OF ABBREVIATIONS

Argonaute

age-related macular degeneration
antisense strand

antisense oligodeoxynucleotide
adenosine triphosphate

Basic Local Alignment Search Tool
branched PEI

Chinese hamster ovarian
drosophila fragile X protein
dynamic light scattering
deoxyribonucleic acid
N-[l-(2,3-dioleyloxy)propyl]-N,N,N-trimethylammonium chloride
double stranded DNA

dsRNA binding domain

dsRNA binding motif

dsRNA binding protein

double stranded RNA

enhanced green fluorescent protein
eukaryotic initiation factor 2a
fluorescein amidite

fluorescein isothiocyanate

fast performance liquid chromatography

xiii

HIV
IF N
IL

LF 2k
LNA
LPEI
LTR
MBP
MDA

Medipal

mRN A
miRNA

NCBI

nt
PACT
PAZ
pDNA
PEI
PEG
PIWI

PKR

human immunodeficiency virus
interferon

interleukin

Lipofectarnine 2000

locked nucleic acid

linear PEI

long terminal repeat

maltose binding protein

melanoma differentiation associated gene
Merlin-Dicer-PACT liaison domain
minimum free energy

messenger RNA

microRNA

National Center for Biotechnology Information
nanoparticle

nucleotide

PKR activator
Piwi-Argonaute-Zwille

plasmid DNA

polyethylenimine

polyethylene glycol

P-element induced wimpy testis

protein kinase R

xiv

PLGA

PLL

PPGL
pp-luc
Pre-miRN A
Fri-miRNA

PTGS

RIG
RISC

RLC

RSV
shRNA
siDNA
siRN A
sod

SS

ssRNA

poly(lactic-co-glycolic acid)
poly-l-lysine

poly(propargyl glycolide)
Photinus pyralis (firefly) luciferase
precursor miRN A

primary W A
post-transcriptional gene silencing
RNAi deficient

RNA helicase A

retinoic acid inducible gene
RNA induced silencing complex
RISC loading complex
ribonucleic acid

ribonuclease

RNA interference
ribonucleoprotein

respiratory syncytial virus
short-hairpin RNA

small interfering DNA

small interfering RNA

Cu, Zn — superoxide dismutase
sense strand

single—stranded RNA

XV

ss-siRNA

STE

TAR

Tat

TBE

TEM

TLR

TRBP

Tudor-SN

v/v

VEGF

VIG

single-stranded siRNA
Salt-Tris-EDTA

trans-activating response element
transactivator of transcription
Tris-boric acid-EDTA
transmission electron microscopy
Toll-like receptor

TAR RNA binding protein
Tudor staphylococcal nuclease
untranslated region

ultraviolet

volume/volume

vascular endothelial growth factor

vasa intronic gene

xvi

CHAPTER 1. INTRODUCTION

1.1 Significance

The flow of information from the DNA in our genomes to messenger RNA
(mRNA), through the process of transcription, and the subsequent translation of protein
fiom that mRNA, is a crucial process carried out in all the cells of our bodies. Disruption
of that process can have severe effects, which can lead to disease. Understanding how
our DNA is maintained and used for mRN A and protein production is therefore essential
for developing methods for disease treatment and prevention. With the successful
completion of the Human Genome Project in 2003, the DNA sequence for every putative
human gene is known (Consortium, 2004; Venter et al., 2001). Current estimates place
the number of protein-coding genes between 20,000-25,000 (Consortium, 2004).
However, the precise function(s) of the proteins encoded by each gene is not yet known.

Accordingly, an extremely powerful research tool would be one by which gene
expression can be specifically inhibited and the effect on cellular pathways observed.
One tool currently being utilized for this purpose is RNA interference. RNA interference
(RNAi) is a natural pathway that mediates sequence specific changes on mRNA levels
through small interfering RNA (siRNA) molecules (Harmon, 2002). The resulting
reduced protein levels are related directly to the Watson-Crick base-pairing between the
siRNA and the complementary target mRN A. For scientists, RNAi is already proving
useful, with synthetic siRNAs being used on a daily basis to target specific genes of
interest. From the resulting changes in protein levels, protein function and interactions

can in many instances be deduced. Additionally, these small RNA molecules are being

pursued as therapeutics, since they could potentially reduce the levels of proteins related
to various diseases; some are even in clinical trials for treatment of diseases such as
macular degeneration (http://www.clinicaltrials. gov).

The considerable investigation of the RNAi pathway since its discovery has led to
an unfortunate conclusion, that not all siRNAs are able to reduce the level of their target
gene. Selecting, testing, and characterizing siRNAs to find an effective sequence can be
a time consuming and costly endeavor. Compound that with the fact that some siRNAs
even have effects on the expression level of unintended mRN As or cause immune
responses (they resemble viruses in many ways), which can mask or preclude their
desired effect, much remains to be learned to realize the full potential of RNAi-based
applications. The work described here helps to improve the understanding of the
mechanism of RNAi, provides additional means for selecting functional siRN A
sequences, and presents a novel method for delivering siRNAs for therapeutic

applications.

1.2 Background

In 1998, evidence was presented suggesting that dsRNA was responsible for
suppression of gene expression in the nematode Caenorhabditis elegans (Fire et al.,
1998). Double-stranded RNAs (dsRN As) targeting well-characterized genes were
injected into adult hermaphrodites, and changes in phenotype, such as body size and
muscle formation, were observed. Small concentrations (only a few molecules) of
dsRNA were required per cell to prevent expression of the target gene, and silencing
endured for several cell divisions (Fire et al., 1998). Prior to this, several laboratories had

observed a similar “co-suppression” effect in plants due to viral replication or transgene

expression (reviewed in (Baulcombe, 2004)). Today, these are classical examples of
post-transcriptional gene silencing (PTGS) that fall under the category of RNA
interference. This phenomenon results in potent and primarily specific gene silencing
that occurs in most eukaryotes (Fire et al., 1998; Harmon, 2002) and is believed to exist
as a mechanism for organisms to sustain genetic integrity by minimizing repetitive DNA
and by preventing deleterious gene expression, such as expression of viral genes
(Conklin, 2003). The profound effect that this pathway has had on our understanding of
basic biology, particularly the role that RNA plays in cell regulation, as well as its
potential for utilization as a research and therapeutic tool, led to the 2006 Nobel Prize in

Physiology or Medicine being awarded to Andrew Fire and Craig Mello for its discovery.

1.3 RNA Interference mechanism

The mechanism of RNAi is composed of two distinct phases (Figure 1-1). In the
initiation phase, long dsRNAs are cleaved by the ribonuclease (RN ase) III enzyme Dicer
into 21-27 nt non-coding RNA molecules called siRNAs (Bernstein et al., 2001; Elbashir
et al., 2001a; Provost et al., 2002). They are then incorporated into the RNA induced
silencing complex (RISC), which in the effector phase uses the siRNA as a template to
bind to complementary mRNAs through Watson-Crick base-pairing. RISC cleaves the
mRNA so that the corresponding protein cannot be translated; the effect is often referred
to as gene silencing or knockdown.

There are several possible sources of dsRNAs greater than ~30 nucleotides (nt) in
length that can be cleaved by Dicer to generate siRNAs. Some dsRNAs are naturally
expressed (W atanabe et al., 2008), such as those produced by viruses or transposons

(reviewed in (Saunders and Barber, 2003; van Rij and Berezikov, 2009)). Others can be

   
 

  
  

RNA Induced
Short interfering Silencmg Complex
RNA (siRNA) (RISC)

Cytoplasm

 

I06

Transcription Sequence Translation .
Specific +
mRNA .

Degradation . .

    
 

. DNA ,
' ' mRNA “Gene Silencing" Protein
Nucleus
Figure 1-1 Mechanism of RNAi.
RNA interference is initiated by the successful delivery of siRNAs to the cell cytoplasm
or by their endogenous expression (1). They are then recognized by the proteins
comprising the RISC loading complex (RLC) (2) and a single strand of the siRNA
duplex, the guide strand, is loaded onto RISC (3). That strand is then used to guide RISC
binding to the target mRNA and subsequent cleavage (4), thereby preventing expression
of the intended protein (5).

exogenously delivered via plasmids that, once transcribed, form short hairpin RNA
(shRNA) that can be similarly processed by Dicer to yield siRNAs (reviewed in (Rossi,
2008)). Conveniently, siRNAs can also be chemically synthesized and delivered to cells,
thus bypassing the need for Dicer processing. A related type of small RNA, known as
microRNA (miRNA), is produced by a similar mechanism (reviewed in (Siomi and
Siomi, 2009)). Primary precursors of miRNAs (pd-miRNA) typically result from
polymerase II transcription products encoded by the genome that form stern-loop
structures. Drosha, another RNase III endonuclease, then processes the pri-miRNAs in
the nucleus into premature miRNA (pre-miRNA) ~65-70 nt long (Lee et al., 2003).
Finally, the pre-miRNA is exported to the cytoplasm (Exportin-S) (Lund et al., 2004)
and, similar to siRNA production, is cleaved by Dicer to yield the mature miRNA.

Much of the current understanding of how RNAi works is based upon studies
done using Drosophila melanogaster. In the fruit fly, the second of two Dicer isofonns
(Dicer-2) is responsible for processing long dsRNAs into siRNAs (Bernstein et al., 2001).
Due to the orientation and processing of the RNA-binding and RNase domains of Dicer-
2, the resulting siRNAs are composed of two complementary strands having 5'-phosphate
groups and 2 nt overhangs on either 3'-end (Bernstein et al., 2001) (Figure 1-2). These
two distinct structural features are critical for their downstream activity yet are
independent of their nucleotide sequence. Dicer-2 and an associated double-stranded
RNA binding protein (dsRBP), R2D2, then jointly interact with the siRNA (Liu et al.,
2003). The heterodimer of Dicer-2/R2D2 positions itself on the siRNA so that R2D2
binds to the more stable end of the duplex (Tomari et al., 2004) while the PAZ domain of

Dicer-2 recognizes the 3'-OH group on the opposing 2 nt overhang (Ma et al., 2004). The

1 Position # 19

Figure 1-2 siRNA structure.

siRNAs are composed of two ~21 nt long complementary RNA strands held together by
Watson-Crick base-pairing. The strands are slightly offset to have 2 nt overhangs on
either 3'-end and each 5'-end is phosphorylated. The top strand here is designated as the
“antisense” or “guide” strand and is complementary to the target mRNA. The bottom
strand is referred to as the “sense” or “passenger” strand.

0 H-3’
PO4-5’

    

positioning of these three molecules, known as the RISC loading complex (RLC), occurs
based on the thermodynamics of the siRNA and the 5'-phosphorylation status (T omari et
al., 2004). The strand bound by Dicer-2 at the 5'-end is used to guide RISC to the target
mRN A and is therefore known as the guide strand. Next, the endonuclease Argonaute
(Ago) 2 associates with the RLC and cleaves the passenger strand in an ATP-dependent
step (Matranga et al., 2005). The passenger strand fiagments are released for further
degradation (Liu et al., 2004) and only the guide strand is preferentially retained in active
RISC (Matranga et al., 2005; Tomari et al., 2004).

Active RISC subsequently uses the selectively incorporated guide strand as a
template to bind mRN As containing regions of Watson-Crick base-pairing
complementarity. It is believed that the first 2-8 nts of the guide strand, known as the
seed region, primarily contribute to target recognition by providing the necessary binding
energy (Haley and Zamore, 2004). Perfect binding between guide strand-loaded RISC
and mRNAs, such as tends to be the case when using an siRNA, leads to Ago2-mediated
cleavage of the cognate mRNA between bases 10 and 11 of the siRNA/mRNA duplex
relative to the 5'-end of the guide strand (Elbashir et al., 2001b; Rand et al., 2005; Song et
al., 2004). However, mismatches between the guide strand and the mRNA, often
encountered when RISC is loaded with miRNAs, frequently can result in translational
repression (Doench and Sharp, 2004). For both mechanisms, translation of the
corresponding protein is inhibited, resulting in the so-called gene silencing that is the
halhnark of RNAi. Several additional proteins can associate with active RISC, such as
Tudor staphylococcal nuclease (T udor-SN), vasa intronic gene (V IG), and the Dros0phila

homolog of fi'agile X-related gene (dFXR) (Caudy and Harmon, 2004; Candy et al., 2003;

Caudy et al., 2002; Ishizuka et al., 2002). However, their role in RNAi is unknown and

appears to be nonessential as only Dicer-2/R2D2/Ag02 is necessary for function.

1.3.1 Human RNAi mechanism

The RNAi mechanism is largely conserved among eukaryotes. Recently,
numerous studies have elucidated many of the details for human RNAi. Human Dicer,
~200kDa, can bind and cleave dsRNAs into siRNAs (MacRae et al., 2007; MacRae et al.,

I 2006; Pellino et al., 2005), as in Drosophila, as well as bind short single—stranded siRNAs
(SS-siRNAs) (Kini and Walton, 2007; Lima et al., 2009). In the cases where siRNAs are
exogenously delivered, if either 5'-end lacks a phosphate, as with chemically-synthesized
siRNAs, a phosphate is added by Clp (W eitzer and Martinez, 2007). The siRNAs then
become part of the RLC with Dicer and the ~35kDa TAR RNA binding protein (TRBP)
(Chendrimada et al., 2005; Gregory et al., 2005; Haase et al., 2005; MacRae et al., 2008).
TRBP is a dsRBP with homology to R2D2 and is hypothesized to perform a similar role
in the RLC (Forstemann et al., 2005; Haase et al., 2005; Murphy et al., 2008). Once
Dicer/TRBP associate with an siRNA, one of four Argonaute proteins joins the complex
(reviewed in (Hutvagner and Sirnard, 2008; Siomi and Siomi, 2009)). Only RISC formed
with Ag02 is capable of cleaving target mRNAs (MacRae et al., 2008; Meister et al.,
2004; Rivas et al., 2005; Vickers et al., 2007). Interestingly, recombinant Ag02 can bind
ss-siRNAs alone and use them to guide target mRNA cleavage (Rivas et al., 2005). As
with Drosophila, several additional proteins can associate with RISC, most likely through
specific interactions with a particular RISC component. For instance, Dicer is known to
interact directly with TRBP and the Protein Kinase R (PKR) activator (PACT) (discussed

more below) (Kok et al., 2007; Laraki et al., 2008). Similarly, Ag02 associates with the

DEAD-box helicase DP103 (Gemin3 and ddx20) and P-body components, among others
(Donker et al., 2007; Hutvagner and Zamore, 2002; Jakymiw et al., 2005; Lian et al.,

2007; Liu et al., 2005a; Liu et al., 2005b; Meister et al., 2005).

1.3.2 dsRBPs involved in and related to RNAi

Double stranded RNA binding proteins (dsRBPs) contribute to numerous critical
cellular pathways, including gene silencing, RNA editing, protein phosphorylation,
transcription, and translation (Chang and Ramos, 2005). By definition, they bind to
dsRNA, typically doing so through dsRNA binding domains (dsRBDs). dsRBDs
maintain a highly-conserved aBBBa (Figure 1-3) structure after folding of ~65-68 amino
acids that can interact with A-form RNA via the 2'-OH and backbone phosphate groups
(Doyle and Jantsch, 2002; Nanduri et al., 1998; Ramos et al., 2000; Ryter and Schultz,
1998; Saunders and Barber, 2003). Additionally, most dsRBDs appear to rely on at least
16-20 base-pairs of dsRNA for recognition (Manche et al., 1992; Ryter and Schultz,
1998; Zheng and Bevilacqua, 2004) since shorter sequences may not allow sufficient
protein/RN A contacts, as illustrated by the crystal structure of a dsRBD from Xenopus
Xlrbpa with a short non-physiological dsRNA comparable in size to an siRNA (Ryter and
Schultz, 1998). However, two reports show that as few as 11 base-pairs (equivalent to
one helical turn of dsRNA) can be sufficient for binding by a single dsRBD (Manche et
al., 1992; Nanduri et al., 1998). Binding is generally acknowledged to be sequence
independent because the major groove, where specific base interactions are mediated, is
narrow and deep and therefore inaccessible (Chang and Ramos, 2005; Fierro-Monti and

Mathews, 2000; Seeman et al., 1976). However, tolerance for disruptions of the dsRNA

TRBP PKR PACT

Figure 1-3 Double stranded RNA binding domain structures of three dsRBPs.

The solution structures for the dsRBD of TRBP (dsRBD 2; Protein Data Bank# 2cpn),
PKR (both dsRBDs; #1qu6), and PACT (#2dix) reveal conserved aBBBa structures that
are believed to mediate contacts with the 2'-OH and phosphate groups of the dsRNA
(reviewed in (Chang and Ramos, 2005)).

helix exist and may in some instances even contribute to binding (Bevilacqua et al., 1998;
Nallagatla et al., 2007). Furthermore, dsRBPs do not tend to bind ssRNA unless they
form secondary structures like hairpin loops (Saunders and Barber, 2003).

A classic source of dsRNA is viral infection, and accordingly, it is no surprise that
many, dsRBPs participate in the innate immune response. More recently noteworthy for
its presence in human RISC, TRBP was originally identified by its ability to bind the
Trans-activating region (TAR) located in the 5'- and 3'-long terminal repeats (LTRs) of
all hmnan immunodeficiency virus (HIV-1) mRNA transcripts (Gatignol et al., 1991).
TAR RNA forms a classic stem-loop secondary structure (Figure 1-4) where the stem,
bulge (nts 23-25), and loop (nts 30-3 5) are important for transactivator of transcription
(Tat) binding and subsequent mRNA transcription (Berkhout and J eang, 1991; Rana and
Jeang, 1999). TRBP facilitates Tat binding, possibly by altering the local structure of
TAR or by mediating binding of other cofactors, and enhances transcription rates, aiding
HIV-1 infection (Erard et al., 1998; Gatignol et al., 1991). TRBP has since been shown
to have oncoprotein-like behavior as its over-expression promotes proliferation
(Benkirane et al., 1997; Eckrnann and Jantsch, 1997). Therefore, TRBP provides a direct
link among several biological pathways, including viral recognition and gene silencing,
by the manner in which it interacts with various types of dsRNA.

Interestingly, a related dsRBP, Protein Kinase R (PKR), can also bind TAR RNA
(with a dissociation constant of ~1-100 nM), but, rather than facilitating HIV-1
expression, it suppresses function (Bevilacqua and Cech, 1996; Kim et al., 2006). It was
therefore not surprising to find that TRBP inhibits the antiviral effect of PKR (Benkirane

et al., 1997). PKR binds dsRNA via its two dsRBDs that work in tandem

ll

UA—u
G_C
A—U
C_G
C_G

A
G_C
A—U
U—A
U—A
G_C
G—U
U—A
C_G
U_G
C_G
G_C

5' G—CB’

Figure 1-4 TAR RNA stem-loop secondary structure.
TRBP was discovered due to its binding of this TAR RNA stem-loop structure that forms
at the 5'- and 3'-ends of HIV-1 transcripts.

12

(Bevilacqua and Cech, 1996; Bevilacqua et al., 1998; Carlson et al., 2003). Upon
binding, PKR homodimerizes and activates by autophosphorylation, initiating a signaling
cascade that culminates in cytokine production (specifically the interferons, IFN-ot and
IFN-B) and a generalized inhibition of protein translation (through eIF2ot) (Cosentino et
al., 1995; Garcia et al., 2006; Patel et al., 1995). These two results contribute to the
innate immune response to viral infection by killing the infected cell in an attempt to
prevent spread of the infection.

It is not quite clear how TRBP and PKR have opposing roles in their response to
dsRNA despite similar means of binding. One theory is that TRBP sequesters the
dsRNA to preclude recognition by PKR. It is also possible that TRBP and PKR bind
simultaneously to a dsRNA (Cosentino et al., 1995). Like PKR, TRBP contains two
dsRBDs at its N-terrninus that are responsible for dsRNA binding, and its affinity for
TAR RNA and other dsRNA appears to be comparable to that observed for dsRNA
binding by PKR, ~1-100 nM (Bevilacqua and Cech, 1996). However, TRBP only
appears to require the second dsRBD for high affinity binding (Daviet et al., 2000). Also
interesting is that TRBP binding to dsRNA appears to be uncooperative; that is, it does
not show increased affinity for longer sequences (Parker et al., 2008). This is in contrast
to PKR and RDE-4 (a dsRBP from C. elegans that participates in siRNA production)
which show a marked difference in affinity for longer dsRNAs (Parker et al., 2008).
TRBP and PKR can also heterodimerize directly through the C-terrninal domain of
TRBP, called the Medipal domain (Daher et al., 2001), as well as heterodimerize with a
third dsRBP, PACT (Laraki et al., 2008). PACT helps to activate PKR (Chendrimada et

al., 2005; Haase et al., 2005; Kok et al., 2007; Li et al., 2006), but, because it shares

13

~42% homology with TRBP, it is hypothesized to assist in siRNA production (Haase et
al., 2005; Kok et al., 2007). Similarity among the dsRBDs of these (Figure 1-3), and a
number of other proteins involved directly in RNAi, illustrates the likely importance of

these domains and their interactions with RNAs in achieving silencing.

1.3.3 Non-specific effects - Immune response and off-target effects of
siRNAs

Given the similarity of TRBP, PKR, and PACT as dsRBPs and their role in
recognition of dsRNA for RNAi or during viral infection, it is not surprising that siRNAs
can also stimulate an immune response (Bridge et al., 2003; Homung et al., 2005; Judge
et al., 2005; Kariko et al., 2004; Kim et al., 2004; Sledz et al., 2003). Other cytoplasmic
proteins, such as the RNA helicases RIG-l and MDA-S, contribute to this effect, by
activating pathways that stimulate the interferons, like PKR does, and other inflammatory
cytokines (Y oneyama et al., 2004). While each of these proteins is predominantly
localized within the cytoplasm, the irnmunostimulatory effect of siRN As is more often
associated with the transmembrane Toll-like Receptors (TLRs). The family of ~10 TLRs
in humans is traditionally responsive to viruses or bacteria components, with each
receptor showing specificity towards a particular type of pathogen (Meylan and Tschopp,
2006). TLR3 is documented for recognizing dsRNA independent of sequence, including
siRNA in some instances (Kariko et al., 2004), while TLR7/8 is more often associated
with binding ssRNAs (Kleinman et al., 2008; Sioud, 2006). However, TLR7/8 have
recently been implicated in initiating an immune response to certain siRNA sequence

motifs (Homung et al., 2005). These TLRs (3/7/8) are not highly expressed on most cell

14

lines and that could explain why early studies with siRNAs in human cell culture did not
detect an immune response (reviewed in (Judge and Maclachlan, 2008)).

In addition to the immune response occasionally stimulated by siRNAs, siRNAs
can have unanticipated effects on many other untargeted genes by being only partially
complementary to mRNAs besides the intended target (F edorov et al., 2006; Jackson et
al., 2003; Scacheri et al., 2004). It appears that target recognition is mediated through
similarity with the “seed” region of the siRNA, the first 2-8 bases, or with the passenger

strand (Jackson et al., 2006; Lai, 2002).

1.4 Application of RNAi

The primary goal of applying RNAi for research or therapeutic purposes, like any
other drug, is to achieve targeted and controllable changes in the cell while minimizing
undesirable effects. Much of the difficulty surrounding the application of siRNAs is due
to the variability in silencing activity of different sequences, where, unfortunately, the
proportion of siRNAs that are successful in repressing their target gene, termed active or
fimctional siRNAs, is low, not unlike what was found for antisense oligonucleotides
(asODNs), another technique for inhibition of gene expression by targeting the mRNA
(Stein, 1998; Vickers et al., 2003). Thus, a common goal of RNAi-related research is to
develop siRNAs that maximize reduction of the target gene at the lowest possible

concentration while avoiding non-specific effects.

1.4.1 Selection of highly active siRNA sequences

The identification of highly active siRNA sequences is complicated by the sheer

size of the human genome. This makes a randomized approach difficult, since early work

15

with RNAi found that randomly selected siRNA activity could vary greatly (Amarzguioui
and Prydz, 2004). Initial strategies for improving the selection focused on eliminating
sequences based on high (>70%) or low (<40%) GC content and stretches of greater than
four consecutive identical bases (e.g., GGGG) (Elbashir et al., 2002). These two
guidelines made siRNA synthesis more efficient but were not based on mechanistic
understanding.

Since then, as the number of siRNAs tested has increased, empirical rules have
been proposed that can be used as part of elaborate computational algorithms that use
tens of parameters to predict which sequences are active and even what their activity will
be (Ge et al., 2005; Lu and Mathews, 2007; Shah et al., 2007; Shao et al., 2007; Vert et
al., 2006). In many cases, these rules merely result fi'om statistical sampling of large
numbers of sequences and have linrited mechanistic implications. More elaborate
selection algorithms have been developed that firrther discriminate the most important
siRNA structural and sequence features.

Perhaps the most important rule in siRNA design, duplex asymmetry, is derived
from the mechanistic understanding that RISC can sense the relative end stabilities of the
siRN A duplex. In theory, both strands of the siRNA may be incorporated into active
RISC, with approximately half of RISCs able to silence the target complementary to the
guide strand and the other half silencing any targets that happen to be complementary to
the passenger strand. However, the strand whose 5'-end is less stably hybridized within
the siRNA duplex becomes preferentially incorporated into active RISC (Khvorova et al.,
2003; Schwarz et al., 2003; Tornari et al., 2004). The result is that a higher proportion of

active RISCs contain the guide strand (for the desired target), leading to more active

16

target silencing. Many of the positional base preferences that have been identified and
are implemented in siRN A design algorithms tend to yield the desired differential
stability between the two ends (J agla et al., 2005; Lu and Mathews, 2007; Reynolds et al.,
2004; Shao et al., 2007 ; Ui-Tei et al., 2004).

Although duplex asymmetry can lead to favorable guide strand selection after
passenger strand cleavage by Ag02, it is possible that the retained guide strand forms
secondary structure that can inhibit Ag02-loading and subsequent target recognition
(Patzel et al., 2005). Rather than experimentally determine guide strand structure, it is
more commonly predicted using separately developed algorithms (discussed in more
detail below). Strands that form little to no stucture are then weighted favorably in the
siRNA selection algorithm (Koberle et al., 2006; Patzel et al., 2005). However, the
impact of guide strand stucture on selecting effective siRNAs is still under debate and is
not incorporated into all current algorithms (Lu and Mathews, 2007).

Once the guide stand is loaded onto Ag02, RISC must locate and bind to
complementary mRNAs. As one might expect, the accessibility of the target mRNA can
have considerable influence on the accompanying level of silencing, as had been found
for asODNs (V ickers et al., 2000; Walton et al., 2002). mRNA tanscripts within the cell
are known to exist in complexes with numerous ribonucleoproteins (RNPs) that are
important for stabilizing the RNA and for protein synthesis. Naturally, these RNPs pose
a hindrance for RISC binding, and vice versa, but not in a way that is at present
predictable. However, accessibility can also be limited by secondary stuctures formed
by the mRNA. Whereas siRNA and miRNA sequences are only ~21 nt long, mRNAs

can be several 1000 nt long and therefore tend to possess significant numbers of

17

intamolecular base-pairs (e.g., the mRNA tanscript encoded by the enhanced green
fluorescent protein (EGFP) plasmid is predicted to have more than 50% of the
nucleotides involved in base-pairs (Gredell et al., 2008)). Several programs exist that
predict RNA stuctures using “nearest-neighbor” fiee energy values experimentally
determined for short RNA sequences (Mathews et al., 1999). One class of program (e.g.,
mfold/UNAfold) minimizes the global free energy of the entire RNA sequence by
extrapolating from the nearest-neighbor values (Markham and Zuker, 2008; Zuker,
2003). A second class of program (e.g., Sfold, RNA Vienna Package) utilizes a statistical
partition function approach based on the Boltzrnan ensemble of all allowable secondary
stuctures, from which the most likely stucture is selected (Ding et al., 2004; Hofacker et
al., 1994). The programs quickly calculate siRNA guide stand stucture (21 nt long), but
the calculation time and memory requirements scale as the sequence length cubed, so
mRN As can require considerably more time and computational capacity (Lu and
Mathews, 2007).

Several reports have used these programs to predict mRNA stuctures and
demonstate a direct impact on siRNA efficacy (Ameres et al., 2007; Bohula et al., 2003;
Brown et al., 2005; Far and Sczakiel, 2003; Lu and Mathews, 2007; Overhoff et al.,
2005; Schubert et al., 2005; Shao et al., 2007; Vickers et al., 2003; Westerhout and
Berkhout, 2007; Yoshinari et al., 2004). Gene silencing decreased when the orientation
or the degree of partial base-pairing of a target constuct was varied for a single siRNA
(Schubert et al., 2005; Westerhout and Berkhout, 2007). "This idea was extended to full-
length tanscripts (ICAM-1 and survivin) where siRNAs targeting inaccessible regions

(i.e., those with extensive secondary stucture) were ten-fold less active than accessible

l8

sites (Overhoff et al., 2005). One particularly intriguing indication of the importance of
target mRNA stucture on RNAi is illustated by the ability of HIV-1 to overcome
silencing by mutating the siRNA target sequence, doing so in a fashion that also alters its
local RNA secondary stucture (Leonard et al., 2008; Westerhout et al., 2005).

Today, the best siRNA selection algorithms rely most heavily on rules that
support incorporation of the appropriate guide stand, based on siRNA asymmetry, and
account for mRNA target stucture effects (Lu and Mathews, 2007; Shao et al., 2007).
When trained with large data sets consisting of results from experiments with hundreds of
different siRNAs, they can predict siRN A sequences for new targets with upwards of
90% accuracy (Lu and Mathews, 2007 ; Shao et al., 2007). Furthermore, attempts are
being made to select against sequences resulting in non-specific effects, including off-
target silencing and immune responses (Fedorov et al., 2006; Jackson et al., 2003;
Reynolds et al., 2006; Scacheri et al., 2004; Sledz et al., 2003). Considerable effort is
also being made by privately held companies to predict siRNAs. It is unknown what the
various proprietary algorithms incorporate, but they are clearly capable of selecting
effective siRNAs and almost certainly include many of the variables discussed above.
Companies are also beginning to offer sets of validated siRNAs for human, mouse, and
rat genomes that have been identified and experimentally validated to possess "high

activity" by some definition.

1.4.2 Specific uses of RNAi

The most common application of RNAi is for studying gene and protein function.
In this situation, one or more genes can be targeted for silencing by one or more siRNAs.

The effect on the cell can then be analyzed to deduce function of the missing protein.

19

Results such as these are published virtually every day and have already provided much
insight into biological function.

However, it is becoming of greater interest to use small RNAs to target genes of
known function to improve the expression of recombinant proteins in plants and modified
mammalian cells (e. g., Chinese hamster ovary (CHO)). Currently, 60-70% of all
recombinant protein pharmaceuticals are produced in mammalian cells (W urm, 2004).
These proteins tend to be of high value and therefore slight gains in output can have vast
economic benefits. Rather than use siRNAs to initiate RNAi, which result in tansient
changes to gene expression, the host genome can be modified to stably produce shRNAs
that permanently alter expression of genes/pathways detrimental to protein production
(Cox et al., 2006; Dodo et al., 2008). The targeted genes tend to lead to increased cell
density and growth rates, and improved protein solubility, stability, and membrane
permeability (Allen et al., 2008; Allen et al., 2004; Kim et al., 2007; Tang et al., 2007;
Zhu and Galili, 2004).

A valuable use of the RNAi pathway is to teat human disease caused by
unregulated protein expression. siRNAs are already being tested in vivo, and several are
entering, or are currently in, clinical trials (http://clinicaltrials.gov). These first
generation therapeutic siRNAs, summarized in Table 1-1, are typically delivered locally
by physical means and target well-studied pathways. Furthermore, many of the siRNAs
are chemically modified to enhance their stability in vivo (discussed below). In some
cases, the method of delivery (discussed below) is as much the focal point of study as the

response to the siRN A itself.

20

 

H cam—em .msoogaonzm «GHQ 830300 «3939303 8535b. 8 ~ A:
_ x238 £8335 ~52 5&3. Bow 89:6 8-...55
u 053. .8235: RE usage 38m 33¢. x50 $-63 mzu
a Lenses 25? Sages 25> 3925 caused 3.32 Baez?
E22986
= gags 3.59 8388089 ass: assume? 33 $5235 «8823:
a 3:333 ~Ecm> Succeed 38a: Beams}. s25 use»??? €8-53 2: :23
a noesfia "6.25 9:08 33%: 832m
B 33385 mews Swansea 33an assume? a? 530m 8.50 $88 cassette
08f— vofio—z begun. woman. «83%qu 3.80m <25»

Ascens£a3§§9=a “3.5 .3259 a 25:... I 2.3.

21

In each of these clinical trials using an siRNA as the therapeutic agent, the
intention is to reduce the expression of a specific gene. As with any therapeutic, the
expected mechanism of action may not always be the actual mechanism. Recently it was
found that an siRN A targeting the vascular endothelial growth factor (VEGF) or its
receptor (for teatrnent of choroidal neovascularization due to wet amyloid macular
degeneration - AMD) could achieve an effect on the target gene by inadvertently
activating TLR3 (Kleinman et al., 2008). The resulting immune stimulation, as
characterized by the induction of IFN-y and IL-12, was principally responsible for the
therapeutic effect, independent of the RNAi pathway. In two other studies, immune
stimulation due to an shRNA was linked to excess guide stand levels, perhaps from
TLR7/8 responsiveness, or fi'om competition with endogenous miRNAs that caused a
saturation of some of the RNAi pathway proteins, particularly Exportin-S (Grimm et al.,
2006; McBride et al., 2008). These results from shRNA mediated silencing firrther
support investigation of the use of siRNA-based RNAi, as siRNAs are not trafficked and
processed in the same manner as shRNAs and miRNAs. Nonetheless, understanding
stategies for mitigating the side-effects of stable in viva silencing will be useful for
development of teatrnents for chronic diseases as well as for establishing cell lines with

constitutively modified gene expression.

1. 4.3 siRNA delivery

While the improvement of algorithms that effectively identify highly functional
siRNAs is a point of emphasis for in vitra applications of RNAi, the main limitation for
in viva work, and consequently usage in a human clinical setting, is delivery. In some

lower organisms, such as worms, dsRNA (siRNA) can be eaten, absorbed, or injected,

22

resulting in highly efficient, systemic delivery (Fire et al., 1998; Harmon, 2002).
However, in mammals, systemic uptake of naked siRNA (siRNA alone with no delivery
agent) is typically poor, as nucleic acids do not freely diffuse across the cellular
membrane. This effect is greatly compounded due to the rapid degradation of siRN As by
nucleases (< 1hr half-life in serum), binding by other factors found in the blood, and by
clearance by the kidneys (Zhang et al., 2007). Although viral methods have the ability to
overcome these limitations by incorporation of silencing constucts into the infected cells
(reviewed in (Manjunath et al., 2009)), viral gene therapy methods still suffer from a
number of safety issues that have caused multiple gene therapy clinical trials to be
abandoned or curtailed. Thus, much effort is being spent developing non-viral delivery
methods for siRN As so that their therapeutic potential can be realized.

In some instances, uptake can be forced through physical methods, including
intavascular injection, ultasound, electoporation, and gene guns (Wolff and Rozema,
2008). In other circumstances, delivery can be accomplished by aerosolizing the siRNA
for administation to the lungs, or by application of siRNA in a topical cream for dermal
delivery (de Fougerolles et al., 2007; Wang et al., 2008). The siRNAs currently in
clinical trials are delivered either by injection into the eye (for age-related macular
degeneration, AMD) or inhalation to the lungs (for respiratory syncytial virus, RSV) (de
Fougerolles, 2008). Unfortunately, these methods are not viable for delivery to deep
tissues or tumors, targets of significant interest for siRN A therapeutics. As an alternative,
both lipid and polymer based reagents are being pursued as delivery vehicles that can be

applied systemically, with the potential for localized targeting, and that may even allow

23

tansport through the blood brain barrier for teatnent of neurological disorders
(Pardridge, 2007).

Due to their phosphate backbone, all nucleic acids are negatively charged.
Accordingly, delivery vehicles typically utilize electostatic interactions between the
nucleic acid backbone and either cationic lipids (lipoplexes) or cationic polymers
(polyplexes). Condensing the siRNA in the vehicle can be achieved through simple
mixing of the siRNA and vehicle in an aqueous buffer and allowing them to self-
assemble or by more complex techniques such as spray drying (Takashima et al., 2007).
The resulting complexes, which can range in size from ~50-300 nm, have been applied in
vitra or in viva. In several of the conditions studied to date, the complex charge tends to
be slightly positive overall, which facilitates contact with the negative charge of the
cellular membrane without causing substantial interactions with blood components that
would lead to complement activation (Bartlett and Davis, 2007a, b). For systemic
delivery, while there is no universally optimal size, it is generally considered that
complexes must be large enough to protect the siRN A fi'om nuclease digestion and renal
clearance (> 10nm) but also small enough (< 70-100nm) to allow access to cells such as
hepatocytes through capillary circulation (Bartlett and Davis, 2007b; Wolff and Rozema,
2008). After association with the extacellular membrane, the complexes are then
endocytosed (Zuhom et al., 2007). The siRNAs must then escape the endoch vesicles
into the cytosol to initiate RNAi.

The taditional stucture of lipid-based delivery reagents is that of a fatty acid
with a cationic head group and a non-polar hydrocarbon tail. Use of this type of

tansfection was first demonstated using DOTMA with plasmids (F elgner et al., 1987).

24

Early RNAi tansfection was performed using lipids developed for delivery of either
plasmids or antisense oligonucleotides. Since then, a variety of proprietary lipid
formulations have been developed that are specifically designed for siRNA tansfection
to cultured cells, such as Lipofectamine RNAiMAX (Invitogen) or siPORT NeoFX
(Applied Biosysterns - Ambion). These reagents can be very effective at delivering
siRN As to diverse cell lines, but, unfortunately, they can be toxic at concentations only
moderately higher than concentations required for effective siRN A delivery; in viva
experiments have also demonstated similar toxicity concerns (Zhang et al., 2007).
While generally not an issue for cultured cell applications, this small therapeutic window
limits the prospects for use of these types of delivery agents for clinical applications.
Recently, a combinatorial approach was used to test lipid-like compounds termed
“lipidoids” to deliver siRNAs, finding that a diverse subset caused extensive silencing in
conditions ranging from cell culture to non-hurnan primate animals (Akinc et al., 2008).
In many cases, these molecules were also found to have acceptable toxicity profiles in
viva. These readily synthesized materials provide an alternative to taditional lipids,
expand the set of available tansfection reagents, and provide insight as to the physical
and chemical characteristics necessary for the most effective lipid vehicles. However,
due to the relatively limited chemical and stuctural diversity in lipid and lipid-like
vehicles, considerably more effort at creating in viva delivery vehicles has been invested
in using polymeric vehicles for nucleic acid delivery, in general, and for delivery of
siRNAs, specifically.

Polymeric vehicles have stong potential for siRN A delivery applications because

they provide protection fi'om nucleases and other serum components and facilitate

25

endocytosis of the siRNA and its release into the cytosol. Of these, polyethylenimine
(PEI) is perhaps the most commonly used and well-studied (Kircheis et al., 2001). At
physiologic pH, the amine groups of PEI are protonated, providing the positive charge
necessary to condense the nucleic acid (Clarnme et al., 2003). Both linear PEI (LPEI)
and branched PEI (BPEI) have shown encouraging results in silencing applications, with
some molecular weights also demonstating delivery of active siRNAs in viva (Urban-
Klein et al., 2005). In several instances, modifications, such as biomolecule addition
(e. g., cholesterol, antibodies/peptides, or aptamers) or crosslinking have been made to
minimize the toxicity associated with the larger molecular weight PEIs and to improve
and target delivery (Swami et al., 2007). A particularly common modification is the
addition of polyethylene glycol (PEG), which partially shields the charges on the polymer
and siRNA, thereby improving membrane interactions, increasing product release, and
decreasing toxicity (Wolff and Rozema, 2008). Interestingly, there appears to be an
optimal amount and length of PEG per polymer, although the combinations studied have
not led to a consensus set of guidelines (Brus et al., 2004; Kunath et al., 2002; Mac at al.,
2006).

Although PEI is perhaps the most commonly studied polymer, two other
considerably more complex systems have shown intriguing in viva results. The first is
called a “Dynamic PolyConjugate” (Rozema et al., 2003; Rozema et al., 2007; Wakefield
et al., 2005). This vehicle contains multiple functional entities to address delivery in a
stepwise manner. In a manner akin to a multi-stage rocket, groups are released from the
conjugate once they have served their purpose, thus preventing interruption of

downstearn events. Starting with a mernbrane-active poly butyl amino vinyl ether

26

(PBAVE) backbone, PEG and N-acetylgalactosamine (for hepatocyte targeting) were
attached via maleamate linkages. The siRNA is also covalently linked to the backbone
via a disulfide bond. The maleamate linkages are reduced in the endosomal vesicles,
exposing the PBAVE, which lyses the endosomal vesicles. The PBAVE-siRNA disulfide
bond is reduced once the complex enters the cytoplasm, thereby protecting the siRNA
from degradation until it is released in the compartment where it can access the RNAi
machinery. Two different siRNAs delivered using this vehicle silenced their target genes
by ~60—80% (Rozema et al., 2007).

A second system is built by self-assembly of an siRNA with a cyclodextrin—
modified PEI, PEG, and tansferrin (for targeting to tumor cells) (Hu-Lieskovan et al.,
2005; Pun et al., 2004). This complex shows low toxicity and effective silencing of the
target gene in a mouse model and in non-human primates. Together, these two polymeric
systems highlight the utility of incorporating factors for targeted delivery and endosomal
release. However, they also illustate the potential complexity of the polymers required

to address the many limitations that restrict delivery of active siRNAs in viva.

1. 4.4 Chemical modification of siRNA

The numerous studies utilizing siRNA underscore the complexity of the various
overlapping pathways that can be initiated by dsRNA and that care must be taken when
eventually progressing to in viva applications. Chemical modification of the siRNA is
one avenue of pursuit to address the problems of nuclease degradation, immune
activation, off-target effects, cell uptake, and pharmacokinetics, essentially aiming to

increase the longevity and specificity of the siRNA in the cellular environment.

27

Coincidentally, many of these problems were previously encountered with asODNs and
thus those results have at least partially guided siRNA development in this regard.

There are limited ways in which the siRNA can be modified. The overall shape
(i.e., overhangs) or length can be changed, where siRNAs only a few nucleotides longer
than 21 base-pairs can sometimes be processed efficiently by Dicer and may facilitate
guide stand loading into RISC (Kim et al., 2005; Rose et al., 2005; Siolas et al., 2005).
Alternatively, chemical modifications can be made to the three key regions of each
nucleotide: the phosphodiester backbone, the ribose ring, or the nucleoside base.
Modifications to the ribose sugar, specifically at the 2'-position, are probably the most
common alterations used for siRN As. The groups frequently added, such as 2'-O-methyl,
2'-fluoro, and locked nucleic acids (LNA) (Carey, 2007) interfere with hydrolysis. In
some cases, they even enhance activity relative to unmodified sequences (Elrnen et al.,
2005; Terrazas and K00], 2009). However, bulkier groups at the 2' position are less well
tolerated and start to have a detrimental effect on silencing potency. A variety of
backbone modifications are available, with phosphorothioate (PS) linkages being
routinely used because they specifically enhance the resistance of the backbone to
cleavage by RN ases. Unfortunately, substantial PS modifications result in increased
cytotoxicity (Carey, 2007; Rana, 2007). Boranophosphate (BO) linkages, while having
been studied less frequently and being limited in scale by synthesis techniques, appear to
offer similar benefits (Corey, 2007). Base modifications are more limited, but can also
improve resistance to nuclease digestion. However, that tends to come at the price of
silencing because many of the modifications reduce the hydrogen bonding and therefore

decrease hybridization affinity. Other modifications that were originally devised to

28

improve the biodistribution of antisense oligonucleotides, such as direct conjugation to
cholesterol, receptor ligands, and tansport peptides, can potentially be applied for
siRNAs as well, provided they do not prevent recognition of the modified siRNA by
RNAi proteins (reviewed in (de Fougerolles et al., 2007)).

Two more recent reports have shown interesting results utilizing modifications to
siRNAs. The passenger stand, when modified with a 5'-O-methyl group, prevented
phosphorylation and decreased RISC activity of that stand (Chen et al., 2008). This
simultaneously had the fortunate result of reducing off-target effects induced by the
passenger stand. A second study found that substituting DNA nucleotides everywhere in
the 5'-third of the siRNA duplex, on both the guide and passenger stands, eliminated off-
target effects without substantial reduction in siRNA activity (Ui-Tei et al., 2008).
Presumably, RNA at the 3'-end of the guide stand is necessary for interactions with
TRBP or Ag02, either for formation of a stable RLC and RISC or for stabilizing
hybridization to the target mRN A; conversely, the DNAzRN A hybrid formed at the 3'-
end of the passenger stand appears not to permit the interactions that are essential for

RNAi.
1.5 Approach and specific aims

The work described here aimed to develop a better understanding of the pathway
of RNA interference in the effort to develop siRNAs for research and therapeutic
purposes. Three aspects associated with RNAi were investigated; the theory was that if
each step in the application of siRNAs could be optimized, then their overall efficacy

would be maximized. The approach was to study, using experimental and computational

29

techniques, three key aspects of RN Ai to improve the rational selection of active siRN A

sequences and their subsequent delivery.

The specific aims of the present study were to:
1. Characterize the effect of secondary stucture within the mRN A site targeted by

the siRNA on the silencing activity of the siRNA.

Algorithms used to identify active siRNAs have historically relied on
thermodynamic asymmetry and other sequence preferences for filtering, while ignoring
contributions from the secondary stucture of the mRNA target site. Here, certain
stuctures predicted by UNAfold (Markham and Zuker, 2008), in which the global fiee
energy was minimized, were found to be more amenable to silencing based on
experimental results with 15 siRNAs and computational results with an additional 533
sequences. Furthermore, guide stand stucture was also observed to correlate with
activity, independent of mRNA stucture. The methods described can readily be

incorporated into improved siRN A selection algorithms.

2. Determine the characteristics of siRNAs that enhance their recognition in the

RNAi pathway by the TAR RNA binding protein (TRBP).
The ability of human RISC to recognize siRNA asyrnmety has been documented

based on empirical evidence from experiments with hundreds of different siRN As.

However, the mechanism for the sensing remains unknown. It is shown here that TRBP

30

alone can perform this function. This result clarifies the role of TRBP in RNAi and

suggests a possible mechanism for loading of the guide stand in RISC.

3. Develop a biocompatible and biodegradable cationic polymer nanoparticle

capable of delivering siRNA into cells grown in vitra.

A novel polymer system generated with “click” chemisty was used to study a
range of variables contributing to siRN A binding, a key step necessary for eventually
achieving cellular uptake and in viva silencing. High affinity binding of the nanoparticles
with siRNA was found to require a combination of primary and secondary amine groups,
and could be enhanced by including 12-16 carbon long alkyl side chains off the polymer
backbone. Other changes in polymer ftmctional groups also altered the affinity.
Ultimately, uptake of the siRNA into cells grown in culture could be facilitated by the

nanoparticles, thus providing a new platform for further development.

31

CHAPTER 2. IMPACT OF TARGET MRNA STRUCTURE
ON SIRNA SILENCING EFFICIENCY

2.1 Abstract

The selection of active siRNAs is generally based on identifying siRNAs with
certain sequence and stuctural properties. However, the efficiency of RNA interference
has also been shown to depend on the stucture of the target mRNA, primarily through
studies using exogenous tanscripts with well-defined secondary stuctures in the vicinity
of the target sequence. While these studies provide a means for examining the impact of
target sequence and stucture independently, the predicted secondary stuctures for these
transcripts are often not reflective of stuctures that form in full-length, native mRNAs
where interactions can occur between relatively remote segments of the mRN As. Here,
using a combination of experimental results and analyses on a large dataset, we
demonstate that the accessibility of certain local target stuctures on the mRN A is an
important determinant in the gene silencing ability of the siRNAs. siRNAs targeting the
enhanced green fluorescent protein were chosen using a minimal siRNA selection
algorithm followed by classification based on the predicted minimum free energy
stuctures of the target tanscripts. Transfection in HeLa and HepG2 cells revealed that
siRNAs targeting regions of the mRNA predicted to have unpaired 5'- and 3'-ends
resulted in greater gene silencing than regions predicted to have other types of secondary
stucture. These results were confirmed by analysis of gene silencing data fi'om
previously published siRNAs, which showed that mRN A target regions unpaired at either
the 5'-end or 3'-end were silenced, on average, ~10% more stongly than target regions

unpaired in the center or primarily paired throughout. We found this effect to be

32

independent of the stucture of the siRN A guide stand. Taken together, these results
suggest minimal requirements for nucleation of hybridization between the siRNA guide
stand and mRNA and that both mRN A and guide stand stucture should be considered

when choosing candidate siRNAs.

33

2.2 Introduction

RNA interference (RN Ai) is a natural phenomenon resulting in patent and
primarily specific gene silencing that occurs in most eukaryotes (Fire et al., 1998;
Harmon, 2002). RNAi is initiated in cells by the presence of either short interfering
RNAs (siRNAs) or microRNAs (miRNAs), which are small non-coding RNA molecules
~21 nucleotides (nt) long (Elbashir et al., 2001a). The guide stand of these small RNAs
is incorporated into the active RNA induced silencing complex (RISC), which then
targets mRN As possessing regions complementary to the guide stand sequence. Upon
hybridization to the target message, RISC prohibits its tanslation, either by cleavage of
the target mRN A when guided by siRNAs, or non-degradative tanslational repression
when guided by miRN As. Utilization of exogenously delivered siRNAs to silence
desired targets by RNAi has become a powerful tool for facilitating disease diagnosis and
teatnent and improving our general understanding of fundamental biological processes
(Harmon, 2002; Mello and Conte, 2004).

Unfortunately, the proportion of siRNAs that are successful in repressing a target
gene is low, not unlike what was found for antisense oligonucleotides (asODNs) (Stein,
1998). As with asODNs, early stategies for choosing siRNAs focused on the sequence
of siRNAs, eliminating sequences based on GC content and stetches of greater than four
consecutive identical bases (e.g., GGGG) (Elbashir et al., 2002). These guidelines made
siRN A synthesis more convenient but were not based on mechanistic understanding.

After identification of siRNAs that would be amenable to synthesis, genomic uniqueness

34

would be verified by BLAST searching. The remaining candidate siRNAs would then be
tested for activity in cell culture to identify the best silencers.

As data fi'om siRN A experiments has accumulated, more elaborate selection
algorithms have been developed that further discriminate the most important siRN A
stuctural and sequence features. One important design rule is based on the relative end
stabilities of the siRN A duplex, with the stand whose 5'-end is more weakly hybridized
incorporated preferentially into RISC (Khvorova et al., 2003; Schwarz et al., 2003;
Tornari et al., 2004). Specific positional base preferences have been identified that tend
to favor the differential stabilities of the two ends (Jagla et al., 2005; Reynolds et al.,
2004; Ui-Tei et al., 2004). Additionally, formation of secondary stucture in the siRN A
guide stand can impair the ability of RISC to interact with its target mRNA (Patzel et al.,
2005), analogous to what was shown with asODN (Mathews et al., 1999; Walton et al.,
1999). More recently, attempts have been made to select against sequences resulting in
non-specific effects, including off-target silencing and immune responses (F edorov et al.,
2006; Jackson et al., 2003; Reynolds et al., 2006; Scacheri et al., 2004; Sledz et al.,
2003)

However, current siRNA selection guidelines have not typically included possible
impacts of the target mRNA stucture on silencing efficiency. It has been shown that
siRNAs can be equally effective when targeting inside the coding region of the mRNA or
the 5'- and 3'- untanslated regions (e.g., (Y oshinari et al., 2004)). Several reports have
used well-defined helices to demonstate that local target stucture has a direct impact on
siRNA efficacy (Ameres et al., 2007; Bohula et al., 2003; Brown et al., 2005; Far and

Sczakiel, 2003; Overhoff et al., 2005; Schubert et al., 2005; Shao et al., 2007; Vickers et

35

al., 2003; Westerhout and Berkhout, 2007; Yoshinari et al., 2004). Gene silencing
decreased when the orientation or the degree of partial base-pairing of a target constuct
was varied for a single siRNA (Schubert et al., 2005; Westerhout and Berkhout, 2007).
This idea was extended to full-length tanscripts (ICAM-1 and survivin) where siRNAs
targeting inaccessible regions were ten-fold less active than accessible sites (Overhoff et
al., 2005). Native mRNA stuctures likely inhibit the interaction of RISC with the target
RNA, again echoing that which was found for target mRNA stucture on asODN function
(Vickers et al., 2000; Walton et al., 2002). One particularly intiguing indication of the
importance of target mRNA stucture on RNAi is illustated by the ability of HIV-1 to
overcome silencing by mutating the siRNA target sequence, doing so in a fashion that
also alters its local RNA secondary stucture (W esterhout et al., 2005).

In this report, we show that the influence of local mRNA target stucture on the
efficiency of siRNA-mediated RNAi is universal, applying to both endogenous and
exogenous tanscripts, across a variety of cell types. In addition, this influence can be
reliably captured through prediction of the minimum free energy secondary stucture of
the full-length target mRNA. The impact of mRNA target stucture on silencing was also
found to be independent of the predicted stucture of the siRNA guide stand, arguing

that multiple stuctural factors should be taken into account when designing siRNAs.

2.3 Results

2. 3.1 siRNA selection and mRNA target structure determination

We sought to establish a relationship between the gene silencing activity of

siRNAs and the stucture of the target mRNA. siRNAs were designed using Ambion's

36

siRNA Target Finder such that siRN A stuctures were only limited to having a 19 nt
duplex with 3'-UU overhangs. No other sequence restrictions were selected. Of the 35
siRNAs returned by the algorithm, 15 were chosen for synthesis based on their GC
content and the predicted stucture of the target region of the mRNA (Table 2-1 and
Figure 2-1). For our experiments, an EGFP reporter gene exhibiting an ~2 hr half-life
was selected. This particular EGFP has been used frequently for RNAi assessment, with
several effective siRNAs in existence (Kim and Rossi, 2003).

In the MFE stucture, 506 out of 846 (59.8%) of the nucleotides were predicted to
exist in intamolecular base-pairs. We selected our 15 designed siRNAs to sample a
number of uniquely stuctured regions. We considered the scenario where the loop
portion of the stem-loop was at the 5'-end, the center, or at the 3'-end of the siRNA target
site within the mRNA (Figures 2-1 and 2-2). These regions were chosen for our
classification as each has been shown to contribute uniquely to the binding and activity of
RISC (Haley and Zamore, 2004), and the stuctures of the ends of guide stands are also
known to impact silencing efficiency (Patzel et al., 2005). For cental loops, we
expanded the region of availability to encompass 4 nt to either side of the cleavage site
because hybridization stability in this region is known to impact silencing efficiency
dramatically (Elbashir et al., 2001b; Martinez and Tuschl, 2004). A loop length of 4
consecutive unpaired nucleotides was chosen based on what has been shown to be
required for nucleation in birnolecular hybridization (e.g., (Hargittai et al., 2004)). For
each of these three target stucture types, we selected siRNAs having low (<50%),
medium (50% to 60%), and high (>60%) GC content. No siRNAs with very low GC

content (<3 0%) were returned by the algorithm, but this fact was ignored as they have

37

Table 2-1 Details for EGFP targeting siRN As experimentally validated in this study.

 

Start End Antisense Sense GC Content (%)
71 89 ACGCUGAACUUGUGGCCGU ACGGCCACAAGUUCAGCGU1 52
81 99 CUCGCCGGACACGCUGAAC GUUCAGCGUGUCCGGCGAG 62
126 144 GAUGAACUUCAGGGUCAGC GCUGACCCUGAAGUUCAUC1 48
159 177 GGGCCAGGGCACGGGCAGC GCUGCCCGUGCCCUGGCCC 76
274 292 UGCGCUCCUGGACGUAGCC GGCUACGUCCAGGAGCGCA 62
306 324 CUUGUAGUUGCCGUCGUCC GGACGACGGCAACUACAAG 52
318 336 CUCGGCGCGGGUCUUGUAG CUACAAGACCCGCGCCGAG 62
396 414 CAGGAUGUUGCCGUCCUCC GGAGGACGGCAACAUCCUG 57
441 459 GAUAUAGACGUUGUGGCUG CAGCCACAACGUCUAUAUC 43
471 489 CUUGAUGCCGUUCUUCUGC GCAGAAGAACGGCAUCAAG 48
495 5 1 3 GUUGUGGC GGAUCUUGAAG CUUCAAGAUCCGCCACAAC 48
501 519 CUCGAUGUUGUGGCGGAUC GAUCCGCCACAACAUCGAG 52
558 576 GCCGUCGCCGAUGGGGGUG CACCCCCAUCGGCGACGGC 71
597 615 CUGGGUGCUCAGGUAGUGG CCACUACCUGAGCACCCAG 57
639 657 CAUGUGAUCGCGCUUCUCG CGAGAAGCGCGAUCACAUG 52

* Start sites are relative to the start codon of the EGFP coding sequence. GC Content is

21 nt.

siRNA sequences obtained from (Kim and Rossi, 2003).

38

71 81 159

“K \e K

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

126

318 396
/\)
495

501

 

 

 

 

 

 

 

 

 

 

 

 

306
471
i
597 639

Figure 2-1 mRN A predicted structures for siRNAs targeting EGFP.

The global mRN A stuctures were predicted (see Appendices — Materials and Methods
for Ch. 2) and the local target site (black line, * denotes 5'-end of site) is shown for each
siRNA. Target start position is indicated for each of the 15 siRNAs. See (Gredell et al.,
2008) for more details.

 

 

 

 

 

 

 

 

 

39

A B

5' 3.

 

 

3' 5.

5' 5':

31/ 3r

Figure 2-2 Local mRN A target structure groupings.

Four local mRN A stuctures were considered when grouping siRN A target regions.
Classification in any group required an mRNA target region (black) that contained four
consecutive unpaired nucleotides (A) at the 5'-end, '5'-loop'; (B) at the 3'-end, '3'-loop';

(C) centered around the siRNA cleavage site, 'cental-loop'; or (D) nowhere along the
target, 'stem'.

 

 

 

40

been shown to be less active (Reynolds et al., 2004). Search results were similarly
limited for very high GC content (>70%) as only two sequences (positions 159 and 558)
were identified. A lack of siRNAs with these GC contents is attributed to the overall GC
content of the mRNA transcript and not necessarily to the search algorithm. To
maximize the unique mRNA structures targeted, we also synthesized an siRNA predicted
to target a fully looped region (pos. 597). Coincidentally, the siRNA at position 159

targeted a fully stemmed region.

2.3.2 Silencing activity of siRNAs targeting the E GFP mRNA

The activity of each siRNA was assessed in HeLa and HepGZ cells by
cotransfection with the reporter EGFP plasmid. These cell lines were chosen because of
their 1) popularity of use and 2) difference in transfection efficiency. No significant
toxicity due to transfection was observed (data not shown). EGFP protein expression was
quantified directly from live cells (Figure 2-3). Of the 15 siRNAs selected, 10 were
effective in silencing the EGFP > 50% and 8 siRNAs were highly effective, resulting in
silencing > 75%. These highly-effective siRNAs appeared to favor 5'- and 3'-target
structures, or structures predicted to be fully looped (i.e., single-stranded). Conversely,
siRNAs targeting central loops, fully hybridized regions, or regions with very high GC
content showed considerably lower silencing activity on average (Figure 2-3, siRNAs

159 and 558).

2. 3.3 Consideration of larger siRNA dataset — data distribution

To support our observation that the structure of the siRN A target region plays a

role in gene silencing activity, we performed our target structure analysis on the siRNA

41

 

 

 

+ -HeLa
-HeoG2 '
1.0- *
$
7’ l
50.84 #
_l
C a I
.9
30.6— I .
9 |
D.
X
uJo4-
g |
8 ' | I
0.2- I
o. . I
U. I ' |
0 I
L“0.0-

T)

J:

C v-OQQQ FFT‘ (Of-l0 V O) N
0.2.8 285:; has ess§a §2 s
GCContent -0+++ 0+ ---o+ 0H3

5' C 3' S L

Figure 2-3 mRN A target structure dependent gene silencing.

siRNAs targeting the EGFP mRNA (see Table 2-1 and Figure 2-1) were cotransfected
into HeLa and HepGZ cells and EGFP fluorescence levels were measured after 24 hrs.
siRNAs are grouped according to the predicted structure (Figure 2-2; 5' — 5'-loop; C —
central loop; 3' —- 3'—loop; S - stem; L — completely unpaired, 'loop') of their target sites
and by GC content ('-' — <50%; '0' — 50-60%; '+' — 60-70%; 'H' — >70%). Values
represent mean i standard deviation for at least eight independent experiments (n28). All
siRNA treatments were significantly different from Control (Student's t-test; p<0.01)
except those denoted by * for HepG2 cells. Comparisons were also made for each
siRNA relative to the sequence giving the smallest change in EGFP expression level for
the corresponding cell line. HeLa: all siRNA treatments are significantly different from
siRNA 159 except those denoted by # (p<0.01). HepGZ: all siRNA treatments are
significantly different from siRNA 558 except those denoted by $ (p<0.01).

42

silencing database from (Shabalina et al., 2006). We analyzed the distribution of data
within this dataset and found that the distribution of siRNAs was heavily skewed to
siRNAs with high activity (low gene expression level) (Figure 24). Further investigation
revealed that this was partially a result of incorporating siRNAs fiom (Reynolds et al.,
2004), where 180 siRNAs targeting every other position of the cyclophilin B and firefly
luciferase were used. Removal of these siRNAs fiom the dataset rendered the data
relatively unifomily distributed, as would be expected for a well-sampled dataset. Our
subsequent analyses were performed with and without the Reynolds data so as to

characterize any uniqueness associated with this particular set of siRNAs.

2. 3.4 Distribution of base-pairs within the mRNA target and the effect on
silencing activity

MFE mRN A secondary structures were predicted for the complete sequence of
each gene, as published by NCBI, in the dataset using UNAFold v. 3.4, with default
settings (Markham and Zuker, 2005; Zuker, 2003). These structures were used to
determine the influence of the total number of base-pairs in the mRNA region targeted by
siRNA. We observed that siRNAs targeted structures ranging from those completely
unpaired (zero base-pairs in the mRNA target region) to those fully paired (19 base-pairs
in the target; Figure 2-5A), with the majority of target regions containing between 10 and
16 base-pairs. The average silencing efficiency tended to decrease as the total number of
base-pairs increased up to 16, beyond which the number of siRNAs in each group is too

limited to assess any significant trend (Figure 2-SB).

43

 

60; —-— Complete Data
~+~— No Reynolds Data

 

 

 

Number of siRNAs
8

 

 

 

I ' fi ' I ' l ' I

o 20 4o 60 so '160'1éo'1io
Gene Expression Level

Figure 2-4 Distribution of siRNAs.

The frequency of siRNAs giving an average gene expression remaining after siRN A
treatment (grouped into bins of 5%; e.g., 0-5, 5-10, etc.) is shown for the siRNAs fi'om
the complete dataset and afier the data from (Reynolds et al., 2004) was removed.

 

100-

 

 

 

 

 

+ Complete Data
-~~+~— No Reynolds Data
80-

(D d

E

g 60"

U)

u—

o

'5 404

.o

E

3

Z 20.

0- ~ '. , . , - , .

 

o 2 4 6 8 '1'0'112t1'4 1'6'1'81—2'0
Total Number of Base-Pairs

 

 

 

 

 

70 - Complete Data
_ - . II No Reynolds Data
0)) 60- l I l
0 J
._l
.5 5‘"
a q
a 40-
g- .
Lu I
a: 30- u
C .
m
0 20-
m 1
a:
E 10.
a)
2 q
0 l I r V l ' l ' l f l

 

 

6'81b12141618
Total Number of Base-Pairs

Figure 2-5 Influence of the number of base-pairs within the target site.

Shown are (A) the frequency of siRNAs targeting mRN A sites with the predicted number
of base-pairings and (B) the effect of the number of base-pairs in the target site on the
average gene expression level remaining after siRNA treatment. For (B), only points
where more than 15 siRNAs were tested are shown. The linear regressions show a
positive correlation. As in Figure 2-4, data are plotted with and without the data from
(Reynolds et al., 2004).

45

2. 3.5 mRNA target loop size and location effects

Since the total number of base-pairs within an mRN A target appeared to influence
the degree of silencing, we hypothesized that the number and location of unpaired
nucleotides would be equally influential in gene silencing. Using the large dataset, we
determined the location of unpaired loops in the mRNA target sites. We considered
scenarios where loops contained 1, 2, 3, 4, or 5 consecutive unpaired nucleotides and
called this the “window size”. The window was walked along the target site (5‘-end of
mRNA target is position 1), and the average silencing activity for those siRNAs that were
unpaired in that window was calculated (Figure 2-6). Even at a window size of l, the
silencing activity tended to be > 5% better at the 5'- and 3'-ends than at the center,
supporting the results we obtained in our experimental system (Figure 2-3). This trend
became more pronounced as the window size increased from 1 to 4 where the difference
in silencing activity fiom ends to center was 8-10%. It is worthwhile to note that 4
available nucleotides has been shown to permit nucleation of nucleic acid hybridization
(Hargittai et al., 2004), supporting our result. Little change was observed increasing from
a window size of 4 to 5 (Figure 2-6). We therefore defined our loop size for structure
classification as four consecutive unpaired nucleotides (W =4) and refined our target

structure matrix (Table 2-2).

46

 

b 01 01
co 0 N 1"
l . l . I . I

 

 

 

Ah A
ON 0)
I.I. .J.

 

Average Gene Expression Level
t

 

 

l ' l ' l 7 l ‘ l r I r I ' l ' l ' I ‘
1 3 5 7 9 11 13 15 17 19
Window Position

Figure 2-6 Effect of window size on gene expression level.

The profiles of the average gene expression level are shown for siRNAs (complete
dataset) that were completely unpaired in the reading window (W=1, 2, 3, 4, or 5
consecutive nucleotides). Because the window size changes, the number of possible
windows changes concomitantly; that is, there are 19 windows of size 1 but only 16
windows of size 4 along a 19 nt long siRNA.

47

038:9? 8e .32 deepens—z com N-N 0BmE 00m 4788 00w§ 05 mo 98% 05 8 0338 03 £8363
0230—052 .meo%0c Bfioomm 05 E 300009 0.53 020300—05: commune: 03300980 0.58 8 v 5:3 Bang 0.83 0003250 0035.

 

(CZ {\z <\Z (<2 wdm 04? mm mm 57: .5 aim .350

Wwo WNW m a mfim v.3. mm v.: 252 Beam

gov 0.30 EA mvn adv mdv mm MN— afiumfi .m

WNW fix? on a: bén NEW 00 mo m—Ih 3.580

5.: vdm om— MMN Q: Yam Vw ~N— WA .m
.5

023.3% 80D mEQSom Sea 02930M Sec 028.3% 8.5 00280—032 0
0 Z 80—9—80 02 303800 02 803800 02 80—380 @2395 we 83003 003025
~0>0A eoamaexm _0>3 nommm0cnxm
0:00 0w80>< 22%... .«o a 0:00 0m80>< 3.2M? we a
35m ease Bfiofim Séa eofiofim

9.20 53 55...» are: <72: 25 852:: <2: .6 525%: a-" use...

48

For targets that contained both a central loop and either a 5'- or 3'-100p, we
classified these as one of the latter. The distribution of the data was similar in each
group, with noticeably less skew when ignoring the Reynolds data (data not shown). The
5'- and 3'-loop structures showed improved silencing activity, while siRNAs targeting
central loops demonstrated reduced fimctionality. Furthermore, stem structures gave
~12% lower silencing efficiency as compared to 3'-100ps. The functionality of siRNAs
within these groups (Table 2-2) was then compared to each remaining group using an
unpaired Student's t-test. The silencing activity of siRNAs in either the 5'- or 3'-groups
was significantly higher than for siRNAs in the center, stem, and other groups (Table 2-
3); the p-value for the combined 5'- and 3'-groups versus all others was 0.0038 and
0.0005 for the Complete Data and No Reynolds Data, respectively, providing statistical

support to our group classifications.

2. 3. 6 siRNA guide strand structure effects

To demonstrate that siRN A guide strand structures were not confounding our
analyses of mRNA local target structures, we determined the MFE secondary structure
for each guide strand and performed the same analysis as described above for the mRN A.
Guide structures of the 5'- and 3'-loop types (defined using W=4) were twice as common
as central loops, with <2% forming stems; no guide strands formed “other” structures
(Table 2-2). Consistent with a previous report (Patzel et al., 2005), guide structures with
5'-loops resulted in improved gene silencing relative to central and 3'-loops (Table 2-2),

as confirmed by Student's t-test (Table 2-4).

49

Table 2-3 Statistical analysis of siRN A activity between groups for mRNA target

 

 

 

structures.
Complete Data

Central 3’ Stem Other All
5' 0.0281 0.4101 0.0100 0.3007 0.0503
Central 0.0472 0.3812 0.1095 0.0853
3' 0.0192 0.3772 0.1039
Stem 0.0589 0.0150
Other 0.2834
5' and 3' 0.0038

No Reynolds Data

Central 3' Stem Other All
5' 0.0867 0.3439 0.0094 0.1321 0.0822
Cental 0.0446 0.2267 0.4410 0.2075
3' 0.0034 0.0752 0.0232
Stem 0.1940 0.0068
Other 0.3028
5' and 3' 0.0005

 

The silencing activities for siRNAs in the group in the left-hand column were compared
to the activities for siRNAs in the group along the top row and the resulting p-value is
shown. Note that the p-value for the 5' versus 3' comparison did not take into account the
fact that ~5% of the siRNAs can be classified as both 5'- and 3'-loops; therefore the data
are not completely independent and thus those p-values are larger than if the groupings
were independent. However, in the comparison of siRN A activities for groups in the left-
hand column versus activities fiom all other siRNAs not in that group (“All” group in the
top row), only independent siRNAs were considered. We also compared siRNAs in
either the 5' or 3' groups together versus “All”. This p-value was lower than either the 5'
versus “All” or the 3' versus “All” p-values because the 5' and 3' groups were not
significantly different from each other, yet they were included in the “All” grouping.

50

Table 24 Statistical analysis of siRNA activity between groups for guide strand

 

 

 

structures.
Complete Data

Central 3' Stem All
5' 0.0058 0.0491 0.1566 0.0029
Central 0.1255 0.3645 0.0363
3' 0.2539 0.3730
Stem 0.2372
5' and 3' 0.0227

No Reynolds Data

Central 3' Stem All
5' 0.0331 0.0718 0.1264 0.0129
Central 0.2347 0.2168 0.1 123
3' 0.1749 0.3105
Stem 0.1619
5' and 3' 0.0583

 

The silencing activities for siRNAs in the group in the left-hand column were compared
to the activities for siRNAs in the group along the top row and the resulting p-value is
shown. Note that the p-value for the 5' versus 3' comparison did not take into account the
fact that ~30% of the siRNAs in the guide strand structure analysis can be classified as
both 5'- and 3'-loops; therefore the data are not completely independent and thus those p-
values are larger than if the groupings were independent. However, in the comparison of
siRN A activities for groups in the left-hand column versus activities fiom all other
siRNAs not in that group (“All” group in the top row), only independent siRNAs were
considered. We also compared siRNAs in either the 5' or 3' groups together versus “All”.
However, the 5' group, and not the 3' group, was significantly different than the “All”
group. "Therefore the significance in the comparison of 5' and 3' versus “All” is due to the
inclusion of the 5' group results and not the 3' group.

51

We next considered a pair-wise comparison of each mRN A structure group versus
each guide structure group (Table 2-5). siRNAs with any combination of 5'- or 3'-loops
in the mRNA or guide structures tended to give the best silencing, with >5% difference in
activity compared to combinations including central, stem, or other groups. siRNAs with
central loops in the guide strand or targeting stems in the mRNA had lower silencing
activities, as was expected. Taken together, these results suggest that inclusion of basic
mRNA secondary structural information, in parallel with siRNA guide strand structure

details, can improve the likelihood of identifying active siRNAs.

2.4 Discussion

Accurately accounting for significant controlling parameters in siRNA design
continues to pose a considerable problem for RNAi applications. Ideally, all mRNAs
targeted for cleavage by RISC would be entirely single-stranded and free from ribosomes
or other bound molecules, allowing for uninhibited hybridization of any siRNA guide
strand to its complementary target. Even ignoring the presence of ribosomes, it has been
shown that the accessibility of the target mRN A through complex secondary structures
can influence RNAi (Ameres et al., 2007; Bohula et al., 2003; Brown et al., 2005; Long
et al., 2007; Overhoff et al., 2005; Schubert et al., 2005; Shao et al., 2007; Vickers et al.,
2003; Westerhout and Berkhout, 2007; Yoshinari etal., 2004). In this work, our goal was
to facilitate incorporation of target mRNA structure into siRNA design algorithms by
investigating which mRNA structure types were most amenable to silencing. We
experimentally showed that specific regions of the target mRNA were more susceptible
to RNAi silencing than others. Furthermore, these results were in agreement with

computational analyses of data available in the literature. The majority of our results (41

52

Table 2-5 Distribution of siRN A functionality for

classification.
Guide Strand
5' Central 3' Stem
35.4 43.2 40.9 18.9
5' 0.4996 0.0141 0.0876
43 24 64 4
41.6 49.9 50.7 56.0
Central 0.0149 0.4147 0.1948
< 31 41 29 1
g . 35.9 49.5 36.7 70.2
3' 0.3969 0.0258 0.1405
64 19 50 4
45.0 52.1 49.9 N/A
Stem 0.0035 0.4659 0.1011
66 23 71 0

each target structure

 

 

 

 

 

 

 

 

 

 

 

 

Target structures were defined when 4 or more consecutive unpaired nucleotides were
present in the specified regions of either the siRNA guide strand or the mRNA target.
The entries in each box represent the average gene expression level (top), the p-value
from a one-sided Student's t-test (middle), and the total number of siRNAs within that
group (bottom) for siRNAs in the Complete Dataset.

53

of 42 genes) are based on the silencing of endogenous mRNAs, thus avoiding any
artifacts that may have arisen by silencing only exogenous or engineered constructs.
Despite the variety of systems and readouts used to obtain it, the data clearly showed that
siRNAs targeting regions unpaired at either the 5'- or 3'-end silenced, on average, 8%
better than siRNAs targeting regions unpaired in the center or without any unpaired
windows. Though this net effect could be argued to be small, our observation is
consistent with a recent report showing an average difference in knockdown of ~14% for
101 shRNAs when accounting for an energy parameter in target accessibility calculations
(Shao et al., 2007). It is noteworthy that in this same report a difference of ~35% was
found when only considering shRNAs that have a favorable duplex asymmetry. While
we did not take into account duplex asymmetry in our analyses, these results strongly
suggest that our differences in average silencing would be improved by removing
siRNAs that are not asymmetric. Regardless, provided that it comes at low
computational and time costs, the additional information gained by using mRNA target
structure predictions is warranted for incorporation into siRN A design algorithms to
enhance the likelihood of selection of active siRN As.

As the number of experimentally tested siRNAs increases, it becomes vitally
important to note the conditions for both design and application of effective and
inefi'ective sequences when reporting results. By doing so, one could greatly enhance the
set of siRNAs available for bioinformatics studies and development of design algorithms
(Matveeva et al., 2007). Algorithms would be similarly assisted by using silencing
results from siRNAs randomly selected for their target, since more elaborate selection

methods may bias sequences towards parameters already shown to be relevant, such as

54

GC content and other particular positional base preferences (Reynolds et al., 2004). We
therefore designed our siRNAs targeting EGFP using an algorithm developed to take
advantage of a simple method for enzymatic siRNA synthesis
(http://www.ambion.com/techlib/misc/siRNA_finder.html). This algorithm only required
that the target site lie immediately downstream of 'AA' dinucleotides, and so all of our
experimental sequences contained UU overhangs (making the entire 21 nt of the siRNA
complementary to the target mRNA). This is certainly a potential source of bias, and
may be one reason why our initial experiments yielded such a high proportion of active
siRNAs (8 out of 15 silenced the EGFP more than 75%). It is noteworthy that the siRNA
targeting a firlly unpaired region resulted in >75% silencing (Figure 2-3; target 597) and
that the siRNAs targeting fully paired regions (targets 159 and 396) yielded only ~20-
40% reduction, an expected trend for the most and least ideal target types, respectively.
Moreover, the total number of base-pairs within the mRNA target site, another metric for
target site stability, showed a normal distribution around 11 base-pairs, with or without
the data fiom (Reynolds et al., 2004) (Figure 2-5A). This suggests that most siRNA
target sites have at least half of the target region sequestered in native structure. As with
the loop and stem targeting siRNAs, siRNAs targeting regions with fewer total base-pairs
tended to reduce gene expression more effectively (Figure 2-5B).

When analyzing data for parametric information, it is important that the dataset is
as unbiased as possible and contains enough data points to draw the appropriate
conclusions. Beginning with the entire database of (Shabalina et al., 2006), we pared it
down to only those siRNAs targeting human genes shorter than 6,000 nts. This

dramatically increased our structural prediction speed while only reducing our dataset by

55

18%, fi'om 653 to 533 siRNAs. After examining the activity profile of the remaining
sequences, we observed that the data were skewed to siRN As that reduced gene
expression levels below 35% (Figure 2-4), due to the Reynolds data. Despite our
concerns about the unequal weighting in the data, inclusion of these data in our analyses
did not significantly alter the results for 5'-, central, or 3'-loops (Figures 2-5 and 2-6,
Tables 2-2 and 2-3). Some effect, though, was seen on the sequences in the “other”
grouping. "Other” refers to those sequences that have loops but not in regions to be
classified in the 5'-loop, 3'-loop, or central class. The average gene expression remaining
for this group increased from 42% to 51% upon removal of the Reynolds data. Similarly,
the analyses implemented in Table 2-3 (and data not shown) revealed no statistically
significant effect from this group. These results indicate that structures that fall into the
“other” class are not as critical for defining silencing efficiency, though this issue is still
up for debate (Katoh and Suzuki, 2007; Schwarz et al., 2006).

Predicted secondary structures for RNA have been determined using a1 gorithrns
such as mfold (Mathews et al., 1999; Zuker, 2003) or the Vienna RNA Package
(Hofacker, 2003) that utilize nearest-neighbor energies to obtain an MFE structure and a
set of suboptimal structures. However, others have suggested that this unnecessarily
assumes that the MFE structure is the most prevalent structure of the mRNA in the cell
(Ding et al., 2004). Instead, an ensemble of foldings encompassing a statistically
significant sampling (~1000 structures) is considered more appropriate (Ding et al.,
2004). This lends itself to a stochastic approach where the probabilities of interactions
can be assessed. Both methods have proven useful in attempting to account for mRN A

target structure in RNAi (Ameres et al., 2007; Heale et al., 2005; Long et al., 2007;

56

Overhoff et al., 2005; Patzel et al., 2005; Schubert et al., 2005; Shao et al., 2007;
Westerhout and Berkhout, 2007). While we did not consider ensembles in this work, for
design purposes, our results (Figure 2-3 and Tables 2-2 and 2-5) and those of several
others (Far and Sczakiel, 2003; Luo and Chang, 2004; Patzel et al., 2005; Schubert et al.,
2005; Westerhout and Berkhout, 2007) show that mRN A MFE structures determined by
mfold alone provide information that is useful in refining the pool of candidates for
selection of active siRNAs.

It is well-established that RNA secondary structure predictions worsen as the
length of the sequence to be folded increases. It is therefore common to predict folded
structures for sequences of a given length, e.g., < 700 nt (Doshi et al., 2004; Mathews et
al., 2004). For predictions of local structures, the folded sequence is generally specified
to encompass a given length (~100 nt) of sequence upstream and downstream of the
target (Heale et al., 2005; Long et al., 2007). This approach, though possibly valuable for
gleaning additional information about the predicted target region structures, would not be
of interest to individuals designing siRNAs due to the computational expense, the number
of folds to analyze, and the lack of a rigorous definition as to the appropriate sequence
length to fold. As such, we did not investigate the use of a segmented approach for
prediction of our target structures.

The extent to which certain regions of the siRNA guide strands, and consequently
the mRNA regions targeted by those strands, influence RN Ai has been somewhat
controversial (Haley and Zamore, 2004; Long et al., 2007; Patzel et al., 2005; Shao et al.,
2007; Westerhout and Berkhout, 2007 ). One report found that the 5'-end of the siRN A

guide strand contributes more to the binding of the target mRNA than the center and 3'-

57

end, which affect the helical geometry of the siRN A/mRN A hybrid (Haley and Zamore,
2004). This conclusion is consistent with our analysis showing a preference for unpaired
5'-ends in the guide strand over both central and 3'-loops (Tables 2—4 and 2-5). Our guide
strand results support those of Patzel and colleagues who found that for guide strands
forming stem-loop structures the 5'-end was more influential than the 3'-end, with either
being more critical than an unpaired central region (Koberle et al., 2006; Patzel et al.,
2005). These results imply that 3'- and 5'-ends of the mRNA regions targeted by the
guide strands are also important. This is further supported by a recent report that
describes accessibility of the 3'-end of the target site as an important determinant of RISC
activity (Ameres et al., 2007). It has also been shown using a reporter construct that a
free 3'-end in the target region improved RNAi silencing (W esterhout and Berkhout,
2007), though only limited structures were considered which perhaps masked the
contributions from the 5'-end. Destabilizing native mRN A structure both upstream and
downstream of the siRNA target sequence can enhance the association rate of RISC as
well (Ameres et al., 2007 ; Brown et al., 2005), which would suggest an improved
likelihood of silencing. Another recent study found that improved silencing occurs at
target sites bordered on both the 5'- and 3'-ends by regions of high AU content (Nielsen et
al., 2007), as these would presumably have relatively weaker native mRNA structure.
However, two other reports reached no such conclusions, instead noting that
siRNA/mRNA nucleation can occur anywhere along the target site (Long et al., 2007).
These papers used a threshold energy term in the analysis of miRNA (Long et al., 2007;
Shao et al., 2007) and siRNA (Shao et al., 2007) binding to structured targets. Afier

nucleation, a second energy parameter regarding helix elongation was required to fully

58

describe the miRNA behavior. This report (Long et al., 2007) also showed that
nucleation of four consecutive unpaired nucleotides gave better correlations to gene
inhibition . This latter result is in agreement with our observations of siRNA initiated
gene silencing (Figure 2-6). Unfortunately, our attempts to describe the energetics of the
siRNA/mRN A interaction using the RNAup algorithm from the Vienna RNA Package
(Muckstein et al., 2006) or the algorithm of (Heale et al., 2005) were unsuccessful
(Figure 2-7). The fact that an effect was observed at both the 5'- and 3'-ends of the target
suggest that the seed site (i.e., the 5'-end) and the relative differential stability between
the 5'- and 3'-ends are not the only important parameters for si/miRNA functionality.
Most likely, it is easier for mRN A targets with four consecutive unpaired nucleotides at
one or both ends to form a stable hybrid with a guide strand also lacking structure at one
or both ends. Once initiated, the dsRNA helix can elongate, possibly facilitated by the
helicase component associated with active RISC (Robb and Rana, 2007).

Using a broad analysis of siRNA-mediated RNAi fi'om the literature, we have
shown that siRNAs targeting mRN As with predicted regions of four consecutive
unpaired nucleotides at either the 5'- or 3'-ends of the target site are more potent for
inducing RNAi-based gene silencing than when the center is unpaired or when the target
site has no unpaired regions. Additionally, these observations are consistent with the
mechanistic understanding of guide strand loading in the RNAi pathway. As these results
were based on predicted MFE structures, they demonstrate the utility of mRN A

secondary structure predictions in enhancing the likelihood of identifying active siRNAs.

59

 

140? Expression= -0.92*Energy + 20.1
120- - -'
:6100- . -"-1 " '
‘ . . . I I l. - .I. I' ' I
5 I '-' - f' ' "
.4 80- -. - "a- "
g - . g‘ I .
.06, 60" ......... ”2.- _ a! .II: _' an...“ . .
9 R“ F fife-99.13:" '
lg- 40“ - a l- I. ' 'I: ‘ziknuxkhf
. '- I"”590“: ' - ."
CD 20- ' . {Iflr‘q‘i "'4 I"'-..-
c ' . . Ht- -' " '
C) ' I I: "v ""- . -
G o ‘ S'IP‘W‘-' " -

 

-45 -40 ' 35 1 -30 ' 25 ' -2'0
Interaction Energy (kcallmol)

 

140'. Expression= 1.29*Accessibility + 42.8

120- . L .
A e I
°\o d I . . . -I-
:100. 1...: --..
g ‘ ' .H- :. ..I~....::VI.
a) , ' ~I . .‘J. I .
‘J 80‘ ' '5- ..I ' I-k.
: . .. :_. .: '4 .
Q 60' - '.' :f.. ..I 'l- =
a .‘ .‘flhal I '::
a ' _ . -- I11:-
o. 40. .firt'fq‘? “ g -
fij ... .5“ . 1.- I-

‘ . I I“.~#I *II.. I

3 20- -=-':'b 3'51' 'L '-
m . I : U: . I I '
(D 0.. I .l..;.:.il -..-*-'

 

 

0.0 ' 32 ' 0:4 ' 0:6 I 0.8
Relative Accessibility

Figure 2-7 siRN A-mRN A interaction predictions.

1.0

(A) The interaction energies were calculated by the RNAup algorithm fi'om the Vienna
RNA Package. The linear regression (grey line) distribution coefficient (R) is 0.1190.
(B) The relative accessibilities of the mRNA were calculated by the algorithm of Heale,
et a1. (2005). The linear regression (grey line) distribution coefficient (R) is 0.0045.

60

CHAPTER 3 RECOGNITION OF SIRNA BY TAR RNA
BINDING PROTEIN (TRBP)

3.1 Abstract

The recognition of small interfering RNAs (siRNAs) by the RNA induced
silencing complex (RISC) and its precursor, the RISC loading complex (RLC), is a key
step in the RNA interference pathway that controls the subsequent sequence-specific
mRNA degradation. In Drosophila, selection of the appropriate guide strand has been
shown to be mediated by the RLC protein R2D2, which senses the relative hybridization
stability between the two ends of the siRNA. A protein with similar function has yet to
be conclusively identified in humans. We show here that human TAR RNA binding
protein (TRBP) alone can bind siRNAs in vitro and sense their asymmetry in the absence
of ATP. We also show that TRBP can bind 21 nt ssRNA, though with far lower affinity
than for double-stranded siRNA, and that the binding reflects the bias observed with the
full siRNA duplex. This suggests that TRBP binding may be both sequence and stability
dependent. A computational analysis of published silencing results supports this
hypothesis. Together these results demonstrate the importance of the siRNA-TRBP

interaction in the formation of active RISC and in siRNA guide strand selection in RN Ai.

61

3.2 Introduction

RNA interference (RNAi) is a means of enacting specific silencing of the
expression of a target gene (Fire et al., 1998; Harmon, 2002). The pathway is initiated
when long dsRN As or pre-miRN As are processed by the RNase III family enzyme Dicer
into 21-27 nt long siRNAs or miRNAs with 5'-phosphates and 3'-dinucleotide overhangs
(Bernstein et al., 2001). The resulting small noncoding RNA is then loaded into the RNA
induced silencing complex (RISC) (Maniataki and Mourelatos, 2005). Typically, RISC
cleaves the target mRNA when using an siRN A to guide sequence complementarity,
whereas miRNAs tend to result in translational inhibition (Siomi and Siomi, 2009).
RISC is minimally composed of Dicer, the TAR RNA binding protein (TRBP), and
Argonaute 2 (A g02), although several other proteins have been shown to associate with
RISC in viva (Chendrimada et al., 2005; Gregory et al., 2005; Haase et al., 2005; Kok et
al., 2007; Lee et al., 2006; MacRae et al., 2008; Robb and Rana, 2007). Dicer appears to
couple dsRNA processing with selection of one strand of the siRNA for loading onto
Ag02 (Kim et al., 2005; Rose et al., 2005; Siolas et al., 2005), the catalytic component of
RISC responsible for degradation of the opposite siRNA strand and the target mRNA
(Liu et al., 2004; Matranga et al., 2005; Meister et al., 2004; Rand et al., 2005; Rivas et
al., 2005; Song et al., 2004). The precise contribution of TRBP to this process has yet to
be established.

TRBP was first identified by its ability to bind the TAR RNA structure present in
human immunodeficiency virus (HIV-1) transcripts (Gatignol et al., 1991). Later, it was

shown that TRBP can inhibit Protein Kinase R (PKR), an important contributor to the

62

innate response to viral infection (Benkirane et al., 1997). Most recently, it was found
that TRBP interacts directly with Dicer, as well as the activator of PKR (PACT), through
the Medipal domain present at the C-terminus of TRBP (Laraki et al., 2008), and that the
interaction is not mediated solely by dsRNA (Haase et al., 2005). A similar interaction is
detected in Drosophila, where Dicer-2 associates with the dsRNA-binding protein
(dsRBP) R2D2 (Liu et al., 2003). Together, the Dicer-2/R2D2 complex senses the
relative thermodynamic asymmetry within an siRNA (T omari et al., 2004). R2D2 binds
to the more stable (dsRNA-like) end, leaving Dicer-2 to bind the opposite end.
Furthermore, binding is enhanced by the presence of a 5'-phosphate on the passenger
strand; hydroxyl groups, such as those present during chemical siRN A synthesis, inhibit
binding (Tornari et al., 2004). As a result of the biased binding, a specific strand of the
siRNA, termed the guide strand, is preferentially loaded into RISC, and the
complementary passenger strand is degraded (Matranga et al., 2005; Rand et al., 2005).
Human RISC has been extensively shown to sense asymmetry between the ends
of the siRNA or miRNA, leading to the preferential incorporation of a single strand as the
guide strand (Gregory et al., 2005; Khvorova et al., 2003; Maniataki and Mourelatos,
2005; Reynolds et al., 2004). This has been accomplished with crude cell extracts and
immunopurified proteins (Gregory et al., 2005; Maniataki and Mourelatos, 2005), and
more recently using only recombinant Dicer, TRBP, and Ag02 in a 1:1 :1 stoichiometry
(MacRae et al., 2008). The effect has also been observed empirically during the analysis
of large siRNA or miRNA datasets (Khvorova et al., 2003; Reynolds et al., 2004), and is
currently the cornerstone for predicting siRNA efficacy (Lu and Mathews, 2007; Shao et

al., 2007). However, unlike in Drosophila, no studies using human cells or proteins have

63

conclusively shown how the asymmetry is detected. Since Ag02 alone cannot bind and
select the siRNA guide strand (Rivas et al., 2005), it suggests that its associated partners
Dicer and/or TRBP are necessary and capable of initial siRNA loading, even though
other factors such as RNA helicase A (RHA) or PACT may contribute to the effect in
vivo (Kok et al., 2007; Lee et al., 2006; Robb and Rana, 2007).

We were interested in characterizing the binding of siRN As by recombinant
TRBP in vitro to determine whether and how it would sense asymmetry in siRNAs. Our
results show that TRBP binds siRN As and not the analogous siDNAs or siDNA/RNA
hybrids. Furthermore, we demonstrate that TRBP alone can sense the asymmetry of an
siRNA duplex, thus establishing a key function of RNAi in humans. Our data also show
that TRBP-siRN A binding may be somewhat terminal sequence dependent, a result
which was supported through analysis of a large-scale silencing dataset. Additionally,
ssRNAs were recognized with a bias that reflected the asymmetry detected in the full

duplex, providing firrther support for some sequence dependence of TRBP-RNA binding.

3.3 Results

To study the contribution of TRBP to siRNA asymmetry sensing in humans, we
expressed and purified recombinant human TRBP as a fusion product with maltose
binding protein (MBP), as well as MBP alone for control (Figure 3-1). Both plasmids
were generously provided by Professor Anne Gatignol (Daviet et al., 2000; Laraki et al.,
2008). Purified MBP and TRBP products were resolved by denaturing gel
electrophoresis and visualized by protein staining or western blotting (Figure 3-1A-C).
Proper function of the TRBP was then characterized by its ability to bind TAR RNA

using a native gel shift assay (Figure 3-1D). Multiple shifted complexes were visible,

64

consistent with previously observed multimerization of TRBP (Cosentino et al., 1995;

Daviet et al., 2000).

3. 3.1 TRBP recognition of small nucleic acids

TRBP has previously been shown to bind siRNAs (Katoh and Suzuki, 2007;
Parker et al., 2008; Ui-Tei et al., 2008). We also found that to be the case in a native gel
shift assay (Figure 3-2, lanes 1 and 2). TRBP formed at least four distinct shifted
complexes with the siRNA, corresponding approximately to complexes containing at
least one siRNA molecule in a complex with one, two, four, or ~six TRBP molecules
(Figure 3-3). As expected, larger complexes became more prominent at higher TRBP
concentrations. Furthermore, although TRBP can bind long ssRNAs (Gatignol et al.,
1993), no binding was observed for the ssRNA corresponding to the antisense strand
(AS) of the duplex siRNA at the TRBP concentration tested (Figure 3-2, lanes 3 and 4).
In addition, no binding was found for an siDNA or the corresponding single-stranded AS
siDNA counterpart (Figure 3-2, lanes 5-8). We next tested DNA-RNA hybrids as these
have been shown to be RNAi-competent (Hohjoh, 2002; Larnberton and Christian, 2003;
Ui-Tei et al., 2008). Somewhat surprisingly, no binding was detected with the
heteroduplexes, either. Repeating the experiment with MBP alone confirmed that any
observed binding was due to TRBP and not non-specific binding by the MBP portion of
the fusion product (Figure 3-4). These results are consistent with dsRBPs, such as TRBP,

PKR, and PACT, requiring A-form, double-stranded regions for binding.

65

 

 

D-
C-
B B"
A-
C q <2 .
M 6° «33’
200---
140-- _
100-
so-T' b
60'“
50---

40 -

Figure 3-1 Characterization of recombinant TAR RNA binding protein (TRBP).

Recombinant TRBP was prepared as a fusion product with maltose binding protein
(MBP) as described (Materials and Methods). (A) Purified MBP or TRBP proteins were
resolved by SDS/PAGE and visualized by staining with Gel Code Blue. (B and C)
Western blots of gels as shown in (A) were performed with antibodies specific for (B)

MBP or (C) TRBP. (D) Gel mobility-shift analysis of TRBP with 32P-labelled TAR
RNA was achieved by incubating 1000 nM of protein with a limiting amount of TAR

RNA (104 cpm; <1.0 nM) and resolving protein-RNA complexes by native PAGE. Four
distinct complexes were visible (bands A-D). M = protein size marker (kDa).

66

 

 

 

 

 

 

V‘ 2 2 2
E < <1: <1:
2 Z Z Z
a: D o: D
\ \ m \ \
w W W U1 m
<
2 <‘r‘ é e i 2
Z Z Z
or a: C) D D c:
— + - + - + - + - + - +
g E.
A-

 

123456789101112

Figure 3-2 TRBP binding2 preferences for nucleic acids.

Native gel shift assay of P- labeled siRNA (lanes 1&2), ssRNA (3&4), siDNA (5&6),
ssDNA (7&8), and DNA/RNA (9&10) or RNA/DNA (ll&12) hybrids with (+) or
without (-) 2000 nM TRBP. Shifted TRBP-siRNA complexes are indicated by A
(monomer of TRBP with siRNA), B (dimer of TRBP), C (tetramer of TRBP), and D
(larger order structure of TRBP).

67

    

12345

 

M123 45

Figure 3-3 TRBP multimerization on siRNA.

Native gel shift assay of a 32P-labeled siRNA with increasing concentrations of TRBP (0
to 2000 nM). The gel was first stained with Gel Code Blue for total protein (left; M =
protein size marker (kDa)) and then used for radio-imaging (right). Shifted TRBP-
siRNA complexes are indicated by A (monomer of TRBP with siRNA), B (dimer of
TRBP), C (tetramer of TRBP), and D (larger order structure of TRBP).

68

 

 

 

 

 

 

 

V)

m < V! U)

< < <

< <2 <2: <

2 Z 2 z

o: D GE 0

\ \ \ \ <
m < W V" 1n 0:
<1: <1: <1: < < < m
z z Z Z Z z <
a: a: D C) D c: t—
— + - + - + - + - + - + — +

I a.“

 

123456789101112-1314

Figure 3-4 No MBP binding of small nucleic acids.

Native gel shift assay as described in Figure 3-2 but with TRBP replaced with MBP. No
shifted complexes are observed for any of the nucleic acids tested, including TAR RNA,
indicating that MBP does not mediate binding for the MBP-TRBP firsion product.

69

3. 3.2 T RBP asymmetry sensing in siRNAs

It has been hypothesized that TRBP senses siRNA asymmetry and therefore
contributes to guide strand selection in RISC. In order to investigate this putative
function of TRBP, we prepared three siRNAs with distinct thermodynamic features. Two
of the siRNAs are known to load into Drosophila RISC based on their strand
thermodynamics (Schwarz et al., 2003) and are consistent with AAG energetics calculated
with existing methods (Hutvagner, 2005) using current nearest-neighbor parameters and a
terminal A:U penalty of 0.5 kcal/mol (Mathews et al., 1999) (see also Appendices —
Materials and Methods for Ch. 3). These sequences correspond to pp-luciferase (pp-luc)
(asymmetric, 'A'; AAG = (AG for the AS 5'-end — AG for the SS 5'-end) = 0.5 kcal/mol)
and human Cu, Zn-superoxide dismutase (sodl) (symmetric, 'S'; AAG = 0.0 kcal/mol).
The third sequence targets EGFP and was found to be moderately active (~60%
knockdown at 33 nM concentration) fi'om our previous work in human HeLa and HepG2
cells but is predicted to be essentially symmetric (EGFP, 'G'; AAG = -0.1 kcal/mol)
(Figure 3-5A) (Gredell et al., 2008). Position 20 of each strand was chemically modified
with a 4-thiouracil to allow for position—specific photocrosslinking (Sontheimer, 1994),
similar to the methods used for characterizing the Drosophila proteins using 5-iodouracil

modified siRNAs (T omari et al., 2004). For each of these sequences, a total of six

different siRNAs were created by alternately hot ([7-32P]-ATP) or cold (ATP) end-

labeling the chemically modified or unmodified single-stranded RNAs, followed by
annealing with the corresponding complementary strand (Figure 3-5B, a-f). The relative

crosslinking efficiency of each siRNA was then determined using recombinant TRBP in a

70

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Asmetric ' 99‘1“: a {Pillmnnnmmlir
AS 5 ' - UCGAAGUAUUCCGCGUACGUG-3 '
SS 3’—AUAGCUUCAUAAGGCGCAUGC -5' bffimnnnmmurg‘l

-82 -8.7

Culplllllllllllllllillifl
Symmetric — sodl

SS 5'- GAGACUUGGGCAAUGUGACUU-3’ dJHHmHHUHHHSI
AS 3

 

’ -UUC CUGAACCCGUUACACUG .‘5 ’ e *p
-10.3 I I -10.3 ‘3

f {Illlnunuummgr

 

 

 

Unknown — EGFP#274 .
fl
AS 5'- UGCGCUCCUGGACGUAGCCUU-B' g LP”HIHHHHHHHP

ss 3 ' - GAGGACCUGCAT cos -5' . T
.114 -113 hti’llllllllllllllillllf

t .___32P

o = 4—thiouracil

 

 

 

 

 

 

 

Figure 3-5 Nucleic acids used.

(A) Three siRN A sequences were used in this study. Two have been documented
previously for their asymmetric (pp-luc; referred to in the text as sequence 'A') or
symmetric (sodl; sequence '8') silencing activities and recognition by Dicer-2/R2D2 in
Drosophila (Schwarz et al., 2003; Tomari et al., 2004). The third was previously used for
EGFP silencing and is expected to be slightly asymmetric based on silencing activity
(Gredell et al., 2008) and AAG calculation with mfold (Zuker, 2003) (EGFP; sequence
'G'). AS denotes the antisense strand and SS denotes the sense strand. Boxed numbers
indicate AG values (in kcal/mol) calculated from the four terminal nearest-neighbor base-
pairs, including the A:U penalty and 3'-overhang contribution (single base stacking
energy). (B) A series of 8 different siRNAs were created for each sequence as described
in the text Appendices— Materials and Methods for Ch. 3. The strand location m (A)

matches the strand location In (B). * denotes strands 5'- labeled with3 2P- ATP; 0 denotes

4-thiouracil modification at position 20; arrows indicate positions destabilized by
sequence mismatches on the bottom-strand.

71

denaturing gel shift assay (Figure 3-6A and D). TRBP showed stronger crosslinking to

the asymmetric 'b' siRNA (Ab) relative to the 'd' (Ad) structure (Figure 3-6A and D;

compare Ab to Ad; p = 1x106) and similarly for the EGFP siRNA (Figure 3-6A and D;

compare Gb to Gd; p = 2x106). However, almost equal crosslinking was detected for the

symmetric siRNA Sb and Sd forms (Figure 3-6A and D; compare Sb to Sd; p = 0.19). As
expected, little signal was returned when radiolabel and crosslinker were not present on

the same strand, showing that crosslinking is specific for the 4-thiouracil positions
(Figure 3-6A and D; compare 'b', 'd', and 'f‘ to 'a', 'c', and 'e‘; p < 2x10-4 for all pair-wise

comparisons). As further confirmation, for all three siRNAs, the crosslinking of 't‘ was
approximately the sum of 'b' and 'd'. Since all siRNAs were bound equally in a native gel
shift assay (Figure 3-6B) and were comparably loaded (Figure 3-6C), we conclude that
recombinant TRBP preferentially crosslinks to the more stably hybridized end of the
siRN A duplex. Similar trends were observed using HepGZ cytoplasmic extracts (Figure
3-7), confirming that this behavior is not purely an in vitro artifact and that TRBP can
detect siRNA asymmetry in the presence of other proteins, both competing and not.

It is critical to note that the crosslinking pattern observed for the EGFP targeting
siRNA was not as expected fiom the AAG calculation, with the asymmetry predicting a
slight bias in favor of 'Gd' vs. 'Gb'. We observed that both the 'A' and 'G' siRNAs
contained an A:U at the AS 5'-end and a G:C at the SS 5'-end, whereas the 'S' siRNA
contained a G:C at both ends. Thus, bias based solely on the first base/base-pair on each

end of the siRNAs would be sufficient to explain the TRBP crosslinking. A similar.

72

d e

 

M
Asymmetric Symmetric EGFP

 

Figure 3-6 TRBP asymmetry sensing.

The siRNAs in Figure 3-5 were tested by gel shifi assay with recombinant TRBP for (A)
crosslinking, (B) native binding, or (C) native loading control, as described in the text
and Appendices — Materials and Methods for Ch. 3. (D) The fraction of siRNA
crosslinked by TRBP was quantified within each lane (fraction crosslinked = crosslinked
signal / (crosslinked signal + uncrosslinked signal) ). Values are average i standard
deviation; n 2 5 for 'A‘ and 'G'; n 2 3 for 'S'. M denotes single-stranded siRNA used as
denaturing size marker.

73

D 0.14-

0.12:
0.10:
0.08:
0.06;

0.04-

Fraction Crosslinked

0.02-

 

0.00

' rir

i

 

 

 

 

 

{-

 

 

 

-I-

 

 

 

 

 

 

 

 

+ #1 l I

 

Figure 3-6 (cont'd).

r-I
I I I l

l
abcde

abcde

ll
fabcdef

 

Asymmetric

Symmetric

74

EGFP

 

Mb d f b d f
Asym. Sym.

 

 

 

 

 

 

 

0.05-

.8 .

E 1

g 0.031

e .

Q 0.02-

C .

{:3 0.01-

9 O 00- l I l

LL . 1 fir I I r r
b d f b d f
Asymmetric Symmetric

Figure 3-7 Asymmetry sensing by human cell extract.

Cytoplasmic extracts from human HepG2 cells (~19 ug/lane) were incubated with siRNA
and crosslinked as in Figure 3-6. Quantification was also as described. Values are
average :t standard deviation; n 2 5.

75

conclusion was recently reached from a computational analysis, finding that one nearest-
neighbor parameter at each end of the siRNA gave the best predictive value for silencing

activity (Lu and Mathews, 2007). We pursued this idea further through our own analysis
of a literature dataset (Shabalina et al., 2006), as we had previously used the dataset when
investigating mRNA target structure effects (Gredell et al., 2008). We sorted the
tabulated siRNA activities according to the nucleotide at the 5'-end of the AS relative to
the nucleotide at the 5'-end of the SS. The most active siRNAs tend to have a U at the 5'-
end of the AS and a G at the 5'-end of the SS (Figure 3-8 and Figure 3-9). Generally, the
order of nucleotide preference was U > A > G > C, for the 5'- nucleotide on the AS, and
G > C > A > U, for the 5'- nucleotide on the SS. These observations agree completely
with those made earlier by Reynolds et al (Reynolds et al., 2004). This is consistent with
what would be required for a bias in stability but also implies a role for sequence,
independent of stability. For instance, a U:A combination (not base-pair) is statistically
more likely to be a good silencer than an A:U combination (Figure 3-8 and Figure 3-9).
More work will be needed to evaluate inconsistencies with nearest-neighbor calculations

and the impact of overhangs on such calculations.

3. 3.3 Altering siRNA asymmetry via mismatches

Since the first and last base-pairs in the siRNA seemed to be critical for
establishing siRNA asymmetry, we next considered the effect the introduction of
sequence mismatches would have on TRBP asymmetry sensing (analogous to (Tornari et
al., 2004) for R2D2). Single nucleotide mismatches were introduced at either position 1
or 19 of the bottom-strand of each siRNA (Figure 3-5; compare 'g' and 'h' to 'b'). The

bottom-strands were cold labeled and then annealed to the hot labeled, chemically-

76

Activity
P

 

 

Figure 3-8 Computational analysis of first nucleotide influence on siRNA activity.

siRNAs from the literature (Shabalina et al., 2006) were sorted according to the
nucleotides at both the AS 5'-end and the SS 5'-end (5'-AS:5‘-SS) and plotted in order of
decreasing mean silencing activity. The small box indicates the mean; the center line, the

median; the lower and upper box lines, the 25th and 75th percentiles; the outer bars, the
95th percentile; and the points, outliers.

77

U:GA:GU:C U:AA:C U:U CzGArA G:CG:GA:U C:CG:AG:UC:A C:U
U:G ~ ‘ ‘

 

A:G ~
U:C -
U:A w
A:C ~
U:U »
CzGl
A:A .
G:C —
G:G r
A:U »
G:C P
GzAi
G:U »

 

 

C:A F
C:U »

 

 

Figure 3-9 Statistical analysis for pair-wise comparison between first nucleotide on
either end of the siRNAs.

A t-test was performed on every possible combination of nucleotide pairings (AS 5'—end :
SS 5'-end) from Figure 3-8. The scale is based on the t-value reported by the t-test.
Roughly, for the sample sizes in the dataset, a t-value > 2.0 indicates significance at the p
< 0.05 level.

78

modified top-strand. The mismatched ends should be destabilized, reducing TRBP
binding and either reducing crosslinking ('g') or increasing crosslinking (11'). Each
siRNA was tested with recombinant TRBP in the gel shift assays (Figure 3-10). For the
asymmetric sequence, a mismatch at the 5'-end of the SS did result in reduced
crosslinking (Figure 3-10A and D; compare Ag to Ab; p = 0.008). However, a mismatch
at the 5'-end did not significantly enhance crosslinking (Figure 3-10D; compare Ah to
Ab; p = 0.49). This may reflect the fact that this asymmetric sequence already has
strongly biased TRBP crosslinking, so increasing the stability bias cannot enhance
binding to the more-stable end. However, for the symmetric and EGFP sequences,
neither mismatch resulted in any statistically significant change in crosslinking (Figure 3-
10D; compare Sg and Sh to Sb, p > 0.31; and Gg and Gh to Gb, p > 0.28), which may be
due to the ~2-fold lower crosslinking compared to the asymmetric sequence despite the
fact that all native binding experiments showed nearly complete binding.

As we saw similar binding but different crosslinking among the three siRNAs, we
wanted to revisit TRBP binding and crosslinking of DNA/RNA hybrids (Figure 3-10).
For all three of the sequences tested, none was appreciably bound (Figure 3-2, lanes 9-12,
where the sequences of the heteroduplexes tested - for binding only - correspond to the
EGFP sequence; also Figure 3-10B, 'i' and 'j', where 'i' and 'j' correspond to 'b' and 'd' in
Figure 3-5, with the bottom- or top-strand of the siRNAs replaced with DNA,
respectively) or crosslinked by TRBP (Figure 3-10A and D, 'i' and 'j'). Since it has also
been suggested recently that 5'-O-methyl modifications can alter guide strand selection
and the propensity for off-target silencing (Chen et al., 2008), we created EGFP siRNAs

with the 5'-O-methyl modification on either end and the 4-thiouracil-modified

79

w«.u§u

 

M55133} A'Ih‘b d: .j g h kl

 

Asymmetric Symmetric EGFP

Figure 3-10 Mismatched siRNAs alter TRBP asymmetry.

The siRNAs were tested by gel shift assay for (A) crosslinking, (B) native binding, or (C)
native loading control, and (D) the fraction of siRNA crosslinked by T RBP was
quantified within each lane, as in Figure 3-6. Values are average i standard deviation; n
23. M denotes single-stranded siRNA used as denaturing size marker.

80

0.14-
0.12;
0.10:
0.08;
0.06:
0.045

:2: ...rrir m

lll r'r_r II I r
ghbdijghbdijghbdkl
Asymmetric Symmetric EGFP

Fraction Crosslinked

 

 

 

 

 

 

Figure 3-10 (cont'd).

81

complementary strand being hot labeled (Gk with the bottom-strand containing the 5'-O-
methyl corresponds to Gg because the 5'-O-methyl groups are only available on T
nucleotides and consequently introduced a C:T mismatch at that end; G1 corresponds to
Gd with the top-strand modified). We expected binding and therefore crosslinking in
both cases to decrease due to the proximity of the 5'-O-methyl group and the 4-thiouracil
modification. However, the methylation had no apparent effect on TRBP binding, and
instead increased the overall amount of crosslinking (Figure 3-10A and D; compare Gk to
Gg, p = 0.026, and G1 to Gd, p = 0.012). Thus, precluding the interaction with TRBP is
likely not the means by which these modifications prevent incorporation of these strands

into RISC.

3. 3.4 TRBP recognition of single-stranded small RNAs

Given that TRBP can sense siRN A asymmetry, and despite its apparent inability
to bind short ssRNA (Figure 3-2), we also wanted to consider whether any other short
ssRNA sequences could be bound or crosslinked by TRBP. We repeated the gel shift
assay with the individually hot labeled, chemically modified ssRNAs from each of the
three siRNA sequences, at increased TRBP concentration (~3-fold higher; 1200 nM)
(Figure 3-11). Intriguingly, the strand favorably crosslinked when part of the asymmetric
siRNAs (Figure 3-6A and D, 'b' for 'A' and 'G') was also preferentially crosslinked here as
a single-strand (Figure 3-11A and D; compare AS to SS for 'A' and 'G'), whereas the
single-strands corresponding to the symmetric siRN A were crosslinked more evenly
(Figure 3-11A and D; compare AS to SS for 'S'). Curiously, the native binding pattern
did not appear to reflect that of the crosslinking gel (Figure 3-11B), which might be

expected for a relatively low affinity interaction that could be altered by the conditions

82

 

     

  

55 A5 A5 SS AS S
Asym. Sym. EGFP

 

Figure 3-11 TRBP recognition of short single-stranded RNAs.

The ssRNAs from each siRNA duplex were tested by gel shift assay for (A) crosslinking,
(B) native binding, or (C) native loading control, and (D) the fraction of siRNA
crosslinked by TRBP was quantified within each lane, as in Figure 3—6. Values are
average i standard deviation; n 24. The values listed below the graph are the ratios
between the AS crosslinking and the SS crosslinking.

83

U

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0307
0.25-
“U
Q)
E 0.20- d,
m
m I:
e a
D 0.02-
= l
0
ca 0.01- i T
in l
0.00 . r . . . ,_.',-_,
AS SS AS SS AS SS
Asymmetric Symmetric EGFP

 

Figure 3-11 (cont'd).

84

‘lr—‘T’TT‘I’I

for electrophoresis. While this supports our contention that terminal sequence matters in
TRBP-RNA interactions, firrther exploration will certainly be required to explain this

result and to ascertain the influence of 5'-end versus 3'-end effects.

3.4 Discussion

RNA interference relies on guide strand complementarity to identify the target for
inhibition. Although the manner by which the guide strand is selected in humans has
been elusive, it is known to be accomplished collectively by the proteins of the RLC:
Dicer, TRBP, and Ag02. Our results show for the first time that human TRBP alone
senses the asymmetry of an siRNA duplex in vitro, a critical step in the guide strand
selection process.

An important step in RLC formation is the initial binding of the double-stranded siRNA

duplex. Dicer and Ag02 appear to be less capable of binding siRNA alone (Kini and

Walton, 2007; Rivas et al., 2005), while TRBP shows a relatively higher affinity (KD ~1-

100nM; data not shown and (Parker et al., 2008)). When in solution, TRBP
homodimerizes (Kok et al., 2007; MacRae et al., 2008), which may contribute to TRBP
forming multimeric complexes on dsRNA even as short as 21 base-pairs (e.g., siRNA;
Figure 3-2 and Figure 3-3) in an uncooperative manner (Parker et al., 2008). This
perhaps indicates a role different from dsRBPs that show improved binding with dsRNA
length, such as C. elegans RDE-4 (Parker et al., 2008). Consistent with what has been
observed for other dsRBPs and with dsRBD function in general (Bevilacqua and Cech,
1996; Saunders and Barber, 2003), TRBP was unable to bind double-stranded siDNA or

RNA/DNA hybrids (Figure 3-2 and 3-10). Recent evidence shows that some DNA

85

substitutions are fairly well tolerated in siRNAs for gene silencing, particularly when
present at the 5'-end of the guide strand (Ui-Tei et al., 2008). Substitutions at the 3'-end,
however, could disrupt silencing, presumably by affecting the TRBP binding site. These
results are in agreement with our observations (Figure 3-2 and 3-10), and it would be
interesting to see how shorter DNA segments spread throughout the siRN A would impact
binding and crosslinking by TRBP.

Though all siRNAs appear to be bound by TRBP with similar high affinity, and
apparently independent of sequence (Figure 3-6B and 2-10B and data not shown), in our
hands, TRBP showed a clear bias in crosslinking, seemingly driven by both sequence and
strand thermodynamics (Figure 3-6). TRBP recognized two siRNAs asymmetrically ('A'
and 'G') and one symmetrically ('S') in our crosslinking gel shift assay (Figure 3-6A and
D). For two of the siRN As tested, our results are consistent with those previously shown
in Drosophila RISC loading (Schwarz et al., 2003) and in crosslinking by Dicer-2/R2D2
(T omari et al., 2004), as well as when using standard techniques to calculate relative
thermodynamic asymmetry (Hutvagner, 2005) (see also Appendices — Materials and
Methods for Ch. 3) (Figure 3-5A). Interestingly, the third sequence ('G') is predicted to
be only minimally asymmetric (AAG = -0.1 kcal/mol), but favoring the opposite end fiom
what was experimentally determined (Figure 3-5 and 3-6, Gb vs. Gd). However, one
might anticipate the experimentally observed asymmetry based on the sequence of the 'G'
duplex (Figure 3-5). The end favorably crosslinked in Gb contains a G:C terminal base-
pair (at position 19 of the AS), whereas the opposite end (pos. 1) is an A:U pair. This is
also true of the 'A' siRNA, but not for the 'S' siRNA (there is a G:C pairing at both ends).

Thus, it would appear that the single base-pair at either end is sufficient for predicting

86

asymmetry (Figure 3-8 and Figure 3-9). A recent computational analysis reached a
similar conclusion that the best predictive results were found when asymmetry only
included the first base-pair of each end (Lu and Mathews, 2007). Furthermore, when
Schwarz et al., replaced either of the end G:C base-pairs in the S sequence with an A:U
pair, the siRNA became asymmetric (Schwarz et al., 2003). It may also be that the
relative thermodynamics are insufficient to predict asymmetry and instead the absolute
energetics may also be important, as has been previously suggested (Schwarz et al.,
2003).

In Drosophila, it was shown that the RLC (Dicer-2/R2D2) initiates unwinding of
the siRNA, containing a small amount of ssRNA relative to siRNA (T omari et al., 2004).
We wondered if TRBP might perform a similar function in humans as it has been shown
to destabilize the TAR RNA secondary structure by base unstacking in the region of the
upper-stern/loop (Erard et al., 1998). Although exceedingly high, non-physiological
TRBP concentrations were required, we detected some low-level ssRNA binding activity
(Figure 3-11A). This low affinity interaction with ssRNA echoes what our group and
others have previously shown for Dicer (Kini and Walton, 2007; Lima et al., 2009).
More surprisingly, perhaps, TRBP preferentially crosslinked to the ssRNA in a fashion
that reflected the biased loading of the strands into active RISC (Figure 3-11B; compare
AS to corresponding SS for each sequence) and the overall level of duplex crosslinking
('A' > 'G' > 'S'). We examined if ssRNA secondary structure was responsible for
differences in binding/crosslinking, but no patterns were discemible for structures
predicted by mfold (Zuker, 2003) (data not shown). Likewise, we cannot rule out that

these effects were merely due to non-specific crosslinking of ssRNA and unbound

87

protein. However, the fact that the ssRNA crosslinking was consistent with that detected
for the duplex seems more than coincidental and will require further testing before a
definitive conclusion can be reached.

Accurately predicting which siRNA strand will be loaded into active RISC,
whether based on the individual end base-pairs, the relative energy differential, or some
other metric, is essential for optimizing siRNA activity. Current algorithms tend to
weigh asymmetry heavily (Lu and Mathews, 2007; Shao et al., 2007). There are several
reports that use mismatches to ensure the desired asymmetry. Interestingly, when we
introduced a single nucleotide mismatch in the asymmetric pp-luc siRN A at the end
preferentially crosslinked by TRBP, we observed a significant decrease in crosslinking
(Figure 3-10, Ag vs. Ab), although the overall asymmetry still favored that end. The
corresponding mismatch at the opposite end did not significantly increase crosslinking
(Figure 3-10, Ah vs. Ab), perhaps indicating that there is a maximum bias beyond which
TRBP crosslinking is not enhanced. Unexpectedly, for the 'S' and 'G' siRNAs, the
corresponding mismatches had no effect on TRBP crosslinking (Figure 3-10). It is
unclear why this was the case since the mismatches in S were previously shown to alter
its loading in Drosophila (Schwarz et al., 2003; Tornari et al., 2004), but it may again
reflect the importance of sequence and relative versus absolute strand thermodynamics.
It could also be that more than one mismatch is needed to alter the asymmetry, as was
recently shown (Geng and Ding, 2008). Other work by our group using human cell
extracts found that a single nucleotide mismatch at the guide strand 5'-end of an siRNA
can cause reduced TRBP binding, with a concomitant reduction in silencing efficacy (H.

K. Kini and S. P. Walton, unpublished results). Or, it may be a difference in the function

88

of the human and Drosophila systems. TRBP alone may not entirely recapitulate in
humans the function of Dicer-2/R2D2 in Drosophila, requiring also Dicer and/or other
components. This is also supported by other work fiom our group indicating that Dicer
binding to an siRNA is improved by incorporation of a terminal mismatch (H. K. Kini
and S. P. Walton, unpublished results).

Several chemical modifications have previously been tested with varying degrees
of influence on activity (Chiu and Rana, 2003; Sipa et al., 2007), but it is unlcnown to
what extent the changes in activity are results of changes in TRBP binding and
asymmetry sensing. We tested one such modification, 5'-O-methylation, shown to
control guide strand selection and eliminate off—target silencing by the modified strand in
humans (Chen et al., 2008). Presumably, the 5'-O-methyl group prevents
phosphorylation by hClp (W eitzer and Martinez, 2007) and subsequently its
incorporation into RISC. In our assay, the 5'-O-methylated siRN As were more
effectively crosslinked than the unmodified versions (Figure 3-10A and D), with no
change in overall TRBP binding (Figure 3-10B). This suggests that 5‘-phosphorylation is
not essential for TRBP asymmetry sensing and that the cellular effects on activity or off-
target silencing are a result of additional factors associated with RISC, such as Dicer for
which it was shown the 5'-O-methyl modification can disrupt binding (Pellino et al.,
2005). It would be interesting to see how other modifications that block/enhance RISC
activity, such as A to 1 RNA editing of pri-miRNA (Kawahara et al., 2007), would affect
TRBP binding and crosslinking as it has been shown for at least one dsRBP (PKR) that
modifications that abrogate activity do not necessarily have an effect on dsRNA binding

(Nallagatla and Bevilacqua, 2008).

89

The interaction(s) and positioning between Dicer/TRBP/AgoZ are critical for
proper loading of the siRNA guide strand onto Ag02 and for functional RNAi activity.
We cannot rule out the contributions of these or other associated proteins, but our results
here show that TRBP alone can bind siRNAs and sense thermodynamic asymmetry.
Although Dicer and Ag02 have not been shown to perform that frmction in RISC, Dicer
may assist in asymmetry sensing, as it appears to do in Drosophila (Tornari et al., 2004).
Quite possibly, Dicer also assists in siRNA unwinding, perhaps through its helicase
domain which was recently shown to be important for processing thermodynamically
unstable shRNAs (Soifer et al., 2008). Furthermore, the interaction between the RNase
111 domain of Dicer and the PIWI domain of Argonautes (T ahbaz et al., 2004) could
facilitate loading of the 5'-end of the guide strand onto Ag02 (Rivas et al., 2005). In
Drosophila, Ag02 is necessary for completely unwinding the siRNA (Okarnura et al.,
2004), which may suggest that in humans Dicer/1‘ RBP can facilitate that step of the
processing. Interestingly, a reduction in either TRBP, Dicer, PACT, or RHA seems to
reduce si/miRN A processing and RISC activity, an effect that may or may not be due to
destabilization of other RISC proteins (Chendrimada et al., 2005; Haase et al., 2005; Kok
et al., 2007; Lee et al., 2006; Maniataki and Mourelatos, 2005; Robb and Rana, 2007).
More experiments will be needed to detail the collective contributions of these proteins to

RNAi and to see if TRBP is the only RISC protein that senses asymmetry.

9O

CHAPTER 4 POLYMERIC NANOPARTICLES FOR SIRNA
DELIVERY

4.1 Abstract

The selection of siRN A sequences is clearly an important factor contributing to
the effectiveness of RNAi applications. But no matter how active a particular siRNA is
during initial trials in cell culture, a major determinant of their eventual widespread
therapeutic use in viva is delivery. A variety of methods are being developed to
accomplish both systemic and localized delivery and typically utilize either lipid- or
polymer-based preparations. Despite the success of some of these methods to deliver in
viva, there remains a general lack of information regarding the lipid/polymer
contributions to nucleic acid binding, nuclease protection, cellular uptake, toxicity, and
release of the siRNA, primarily due to vast differences in the systems being developed.
Furthermore, the existence of diverse experimental data published thus far can at least in
part be attributed to the difficulty of modifying many of the lipids/polymers.

The work described here was aimed at developing a novel biocompatible and
biodegradable cationic polymer nanoparticle (NP) capable of delivering siRNA into cells
grown in vitra, with potential use later for in viva applications. The polymer developed
by our collaborators was readily modified by standard “clic ” chemistry methods, which
allowed for a systematic study of variables contributing to siRNA binding. Strong
siRNA binding was found to require an NP containing a combination of primary and
secondary amine groups to facilitate electrostatic interactions. Inclusion of alkyl chains
enhanced binding, perhaps by causing vesicle-like formation. Consistent with its known

charge shielding effects, polyethylene glycol (PEG) could be included to reduce overall

91

binding affinity, and is expected to be important for nuclease protection. Ultimately, the
NPs were capable of binding and then delivering siRNAs to cells, but were unable to
initiate RNAi. Accordingly, subsequent studies will be needed to investigate how
alterations to the polymer affect cellular uptake and siRNA release without greatly
affecting NP/siRNA complex formation so as to maximize the silencing potential of this

siRN A delivery platform.

Author contributions:

The work described in this chapter was performed in collaboration with Professor
Gregory Baker fiom the Michigan State University Department of Chemistry, his
graduate student Erin Vogel, and Professors Christina Chan and S. Patrick Walton from
the Michigan State University Department of Chemical Engineering and Materials
Science.

We all contributed to the project conception, experimental design, and data
analysis. Erin was responsible for synthesis of the polymers tested and for
characterization of NP/nucleic acid complexes by transmission electron microscopy and
dynamic light scattering. I performed the gel shift assays and curve fit analyses to
determine NP binding parameters, with the assistance of several undergraduate students
(Sophie Carrel], Dan Desantis, and Jorge Fontes), as well as all cell treatment
experiments. The chapter itself was written entirely by me with minor editing by

Professor Walton.

92

4.2 Introduction

The therapeutic potential of siRNAs greatly relies on being able to overcome the,
natural barrier of human cells to foreign material, in particular highly charged species like
nucleic acids. Nucleic acids do not freely diffuse through the cellular membrane, are
rapidly degraded by nucleases, bound by other factors found in the blood, and cleared by
the kidneys (Zhang et al., 2007). The intentional delivery of dsRNA into human cells
also presents a unique challenge because of the potential for stimulating an immune
response; viral methods for gene therapy can overcome some of these limitations, yet
they still suffer from a number of safety issues. Instead, much effort is being spent
developing non-viral delivery methods for siRNAs so that their therapeutic potential can
be realized.

As an alternative, both lipid and polymer based reagents are being pursued as
delivery vehicles that can be applied systemically, with the potential for localized
targeting, and that may even allow transport through the blood brain barrier for treatment
of neurological disorders (Pardridge, 2007). However, lipid-based delivery reagents
(reviewed in (Tseng et al., 2009)) present their own drawbacks, such as potentially high
in viva toxicity (erang et al., 2007) and a relatively limited chemical and structural
diversity. Therefore, considerably more effort has been invested in creating polymeric
vehicles for nucleic acid delivery, in general, and for delivery of siRNAs, specifically
(reviewed in (de Martimprey et al., 2009)).

Wide varieties of polymeric vehicles have been synthesized by various chemical

means, but, as with viral methods and lipid carriers, are ultimately restricted based on

93

er

!’”.fi.F—'. '. area an.

their biostability and biocompatibility. For many of the polymers traditionally used for
nucleic acid delivery, such as PEI and PLL, biostability and biocompatibility can be
improved by attaching groups like PEG (for charge shielding and to facilitate membrane
interactions) or other biomolecules (e.g., cholesterol) through standard chemical means
(Kircheis et al., 2001; Swami et al., 2007; Urban-Klein et al., 2005; Wolff and Rozema,
2008). However, the steps involved in modifying these polymers can require diverse
reaction conditions and solvents.

A new method for post-polymerization modification was proposed in 1999 by Dr.
Barry Sharpless' group that was aimed at mimicking biology by using a series of highly
specific, high yield reactions that can be performed under relatively benign conditions
and require little in the way of product isolation (Kalb et al., 2001; Sharpless and Kolb,
1999). Today, these reactions, known as “click” chemistry, have become a powerful tool
for pharmaceutical applications due to the ease wiflr which they are performed. Perhaps
the most prominent reaction meeting the "click" criteria is the Huisgen 1,3-dipolar
cycloaddition of azides and terminal alkynes (reviewed in (Hein et al., 2008)). This
reaction has been implemented by our collaborators (Jiang et al., 2008) in the
development of a novel polymer system that resembles poly(lactic-co-glycolic acid)
(PLGA), another polymer frequently used for nucleic acid delivery (de Martimprey et al.,
2009). Both PLGA, and this polymer, poly(propargylglycolide) (PPGL), . are
biocompatible and biodegradable (into lactic acid derivatives), thereby lending
themselves to a biological setting. However, unlike PLGA, PPGL can be readily

modified by "clic " cycloaddition reactions (through the formation of 1,2,3-triazoles) to

94

create new materials (Figure 4-1A) with potentially enhanced properties (Jiang et al.,
2008).

These characteristics make PPGL an ideal candidate for the encapsulation of
small molecules, including siRNA, for their delivery as therapeutics in _viva.
Furthermore, because the azide group is relatively easily added to many of the moieties
already being included in other complex polymer systems (Figure 4-1B), a family of
related vehicles can be created that incorporates nearly limitless modifications. The final
structure when the azide groups are added to the propargyl backbone is roughly brush-
like, with the azide groups forming the bristles of the brush (Figure 4—1C). Accordingly,
with this polymer backbone it is possible to systematically investigate the effect various

moieties have on siRNA vehicle design.

4.3 Results and Discussion

Our approach was to begin developing the PPGL polymer system developed in
Prof. Baker's laboratory (Jiang et al., 2008) to deliver siRNA to cells grown in culture by
first systematically studying the impact of functional group additions to the polymer
backbone, in particular the influence the various groups had on NP binding of siRNA.
Description of the polymer syntheses and other experimental details can be found in
Appendices — Materials and Methods for Ch.4. Future studies to be performed by others
will aim to characterize and engineer other important parameters, such as nuclease
protection, cellular uptake, cytoplasmic release of the siRN A (endosomal escape), and
activity in RNAi.

We first determined whether a preliminary version of the polymer (unmodified

PPGL) could enter the cells and if so, what was its cellular distribution. A rhodamine

95

unfi. a
..

-1.

/
o / 2
R ‘N3
R\ o 01H ,
O o x catalytic Cu'

é
RAN;
propargylglycolide
B
N3/W\NH2
a 0 /\/N3
4 3 H 2
PEG Amines
(charge shielding) (electrostatic interaction)
C
Amine-containing
Moiety
PPGL
PEG <_Backbone
Alkyl_>
Chains—>

Figure 4-1 Nanoparticle synthesis details.

(A) Nanoparticles were created by adding functional groups (B) to the poly(propargyl
glycolide) (PPGL) backbone. Amine-containing moieties or polyethylene glycol (PEG)
provide positive charge and charge shielding, respectively. Additional types of groups
can be added, as listed in Table 4-1 and shown in Figure 4-4. (C) A sample nanoparticle
is depicted. Attached to the polymer backbone is a PEG tail (lefi end), and amine-
containing moieties and alkyl chains dispersed randomly along the backbone. The
backbone and sidechains are flexible and relatively free to organize into NPs in solution.

96

 

fluorescent group was attached to the polymer and the solution was administered to
human HeLa cells for 24 hrs. Fluorescence microscopy confirmed that the cells appeared
healthy and were indeed internalizing the polymer into the cytoplasm (Figure 4-2A).
Most polymers typically pursued for siRNA delivery form complexes with the
siRNA through electrostatic interactions with the anionic nucleic acid backbone.
Frequently, the siRNA and vehicle are simply mixed in an aqueous buffer and allowed to
self-assemble to form condensed complexes. We therefore considered whether the NP in
question could form a stable complex with nucleic acid as that is the first step in delivery.
Rather than use siRNA to establish binding characteristics, we instead assessed binding
with a dsDNA molecule comparable in length to an siRNA and containing 2 nt 3'-
overhangs (siDNA). The charges for the two types of nucleic acid are the same, albeit
with slightly different organization along the backbone because of the difference in helix
geometry, but the siDNA is approximately 1/ 10th or less the cost of the siRNA. The
polymer and siDNA were mixed in water at room temperature for 15 min and then
subjected to agarose gel electrophoresis to separate bound and unbound nucleic acid (gel
shift assay). No complex formation was observed for this NP (data not shown).
However, shifted complexes were visualized with this method using several of the
polymers discussed above, including LPEI, BPEI, and the lipid solution Lipofectarnine
2000 (LF2k; Invitrogen), indicating that this assay could be used to screen alternate
polymer chemistries (data not shown). The gel shift assay has also been applied by
others for testing polymer/plasmid DNA or polymer/siRNA binding characteristics with

similar success (Bartlett and Davis, 2007b; Jeong and Park, 2002; Mann et al., 2008).

97

 

 

Figure 4.2 Visualization of nanoparticle delivery into cell culture by fluorescence
microscopy.

(A) Images show uptake of rhodamine tagged NP (1.6 rig/ml) into HeLa cells. Top row:
10x mag. Bottom row: 40x mag. Left column: fluorescence. Right column: phase
contrast. (B) Delivery of Dy547-tagged siRNA (30 nM; red fluorescence; upper left) by
untagged NP (1 mg/ml) into H1299 cells stably expressing the Enhanced Green
Fluorescent Protein (EGFP; green fluorescence; upper right). The cells tolerated the
NP/siRNA and remained healthy (phase contrast; bottom center). High EGFP levels
indicate no silencing due to siRNA delivery. All images 10x mag.

98

 

We hypothesized that the failure of the preliminary NP to bind siDNA was due to
little or no electrostatic interaction between the NP and siDNA. Subsequently, several
new NPs were synthesized to investigate the effects on binding of 1) the different types
and amounts of amine groups that could be added to the backbone; 2) the location and
length of polyethylene glycol (PEG; for its charge shielding and membrane interaction
effects); and 3) the addition of alkyl chains dispersed among the amine groups (for
facilitating potential vesicle-like formation). Each NP was then tested by gel shift assay
(Figure 4-3A). The fiaction of siDNA bound, calculated from the fluorescence signal for
the unbound siDNA, was plotted versus NP concentration (Figure 4—3B) and the resulting
binding data were regressed according to the Hill equation for describing nonlinear dose-

response behavior (Goutelle et al., 2008)

[NP]"
K " + [NP]"

to determine the binding constant (K, in rig/ml) and cooperativity of the binding

 

F ractianBaund =

interaction (n) for each chemistry (Table 4-1 and Figure 4-4). It is important to note that
the parameters K and n of the Hill Equation have physical meaning in only limited
situations (Goutelle et al., 2008; Weiss, 1997), which may or may not be applicable here.
The equation is used only as a means to compare binding between different polymer
chemistries without having prior knowledge of the mechanism of interaction between NP

and siDNA.

99

 

..
PEG115PG45(A17O:C1030) 1PG150(A12702C1030) l

 

 

 

 

PEG115PG45(A1270:C1030) LPEI
B 1.2-
R"2= 0.96
1-0- k = 1223 :44 ' '
' n = 2.961 0.38

"C

C

3

O

in

C

.9

U

9

LL

 

 

 

10 .......1.60 . .....1.(.).OO .
NP Concentration (pg/ml)

Figure 4-3 Nanoparticle complex formation with siDNA (analog of siRNA).

(A) siDNA was incubated with decreasing concentrations of NP (or control polymer,
linear polyethylenimine; LPEI) for 15 min and separated by gel electrophoresis. (B)
Sample fit of fraction bound versus NP concentration fitted with the Hill equation. K is
the binding constant (pg/ml) and n is the cooperativity. The nomenclature corresponds to
the structures described in Table 4-1 and depicted in Figures 4-1 and 4-4.

83: A 3868232265 35>
25.5 A 3863342305265 35>

 

 

Good“ A AONO floowgvmvwmm g Hmvmm $.=H>
See A 38882322 H0.5 55>
mo 55 2 a: evoaooefifiegwomm 35>
one 85 a we 38535530565 55>
e3 21 2 n: 8888232252 55>
3o 9: 3 3m 35685332on 85>
So and SN 2.8 emx25<V$om§omm 85>
:5 55 2: we: 5x85332265 35>
moo as 2 2K eMSoQSSRomwomm 55>
25.5 A 356853305265 25>
83: A ASSoSSSSome 65 25>
56 m3 : 2N emoaoowSSSommzoE 25>
85 a; 2 m5 356853305265 25>
Ed :5 mm e2. 338253022 e5>
”no 23. a. £5 GMSoESSSomEoE 55
83: A 3883332265 55>

-\+ a 4+ €535
nougem =5 usefim 0.52

$33 5:... .0» .3 605—532. 3 <25» new 833.3123: me A5 haze-20:89 6.8 A5 saw—3.5m wig—um— Tv o-Aah

101

s~5£2$232 6mm

 

 

go 8.0 0m 8m 25>
3o :2 a 22 emvoowmimvgwmmm 55>
26 NE 2 8m 386252532265 25>
go 55 cm N5 Amcaeomagsomwomm 25>
26 m3 2 2 38685323265 «55>
25 So am am 5552232265 :5>
956— A AomXOBN _ <vmv0mw0mm QWSS
$3 2: 3 mm 33588252255 55>
86 85 2 m: 38035082322”ch 85>
mg 3.5 3 n2 38682320505 25>
25.5 A 3868322on 55>
25.5 A 38685305on 35>
:5 who EN 33 555329305 25>
mod 26 2. $3 5x82320565 $55
:.o a; 3 m2 Biafifimvomwcmm 25>
COCA: A AGNO ~ Uovam— y Uvmnmvm ~m.~=>
2d ”2 2 N5 38882520585 35>
:5 N3 E 26 3N2082$22265 35>
4+ : -\+ 5.335
836m :5 ofiofim 0:32

A9233 TV 033!

102

 

 

 

.oaata EwEO 300.832 mama: cougvm Em 06 5.3» 8% E5 Em mo mafia 3.30 \3 338.893

203 3+v $8.8 wfiucommotoo :05 98 A5 53380908 .03 3838 wfiwfin 2:. .Amvd; £3 32838 333%: 33 080:8
05 898 £033 E 63863 05 9 323 0.53 35on 3:03me 80565 088 E “m5 082 .On 3 Buofié mm onU ”NLXBSHS
8% 03658 wfifiafifi 2: mafia anew wa§:8-:on.ao mo A595: 25 2: mm 0 £83 380:3 3388 033126 2: .«o o\¢N mam—E
anew magnooéfifia mo 25 05 m_ < “am 62393» 18.3905 mo 3:3 “83.— mo .8982: 598— 88303 05 mm a AOm—mv 392w
0835038 mo 33: “8%.. mo .8383: 05 mm a 80:3 A. . SON<VpOmaOmm “mEBozom 2.: 8 wEEoooa Yv earn E 8505. anew

on... 8m 2330:0an 3.58m 05 Bo=£ 850:5 :2; .3Nmmofigm 0.83 >05 £033 5 .520 06 3 wSEooom 38m: 203 mm Z BF

 

 

8.0 v3 3 QR BE
93 w: and SN 35.:
Ed 8; who ta 35.5
25 .85 Sod $8 3m
mod 9o mm «3 GSUONSUENZEEMWBQ 225»
:.o 3o 3 gm e~x§ovofl08823255 ~:.E>
m; a: 3 of 38582332on 3:5,
86 m; E mmfi emxsmzvmvomaomm 35>
one 2: 5. q: emfiomomzusfiseg 35>
.\+ a -\+ €31;

magnum am ofiofim onaz

.5223 3. oz;

103

1:3: »<»)%
22%; W

(128)
A2 WN/
(156) H
A3 WN/
(170) I
| /
A4 ‘fiA/N\
(223) Br
9
A12 H

A122 PEVNMNMNHZ
H

(214) H
G NH
(170) AWNJLNHz
H

Figure 4-4 Nanoparticle functional groups added to the propargyl glycolide
backbone.

The NPs listed in Table 4-1 (and data not shown) were synthesized by incorporating these
groups at the 'R' position (Figure 4-1A) or by reacting with the alkyne positions (to yield
the 'R2' groups). PEG = polyethyleneglycol. PEI = polyethylenimine. D = dabcyl. R =
rhodamine.

104

 

1%

Ex

 

CL
(154)

C4
(99)

C10
(183)

C16
(267)

AP
(242)

CA
(201)

C81t8
(293)

C1EO4C6

(333)

C1E03C12

(373)

(604)

Figure 4-4 (cont'd).

h’VW\/\/\/\/\

 

105

W
J

l
3

 

 

It was clear from the nearly 100 different polymers tested by gel shift assay that
the type and amount of amine-containing groups attached to the polymer backbone
significantly affected siDNA binding characteristics (Table 4-1). NPs containing no
amine groups (no A) or only primary (A1), secondary (A2), tertiary (A3), or a
combination of primary and secondary amine-containing groups on separate chains
(A1A2), showed no detectable binding of siDNA up to NP concentrations of 10 mg/ml
(Table 4-1). Interestingly, when primary and secondary groups were included on the
same chain (A12), significant siDNA binding was observed (e. g., Table 4-1; compare
VIII.14, K = 236 pg/ml, to VIII.43-VIII.46, K > 10,000 ug/ml). Moreover, binding was
further enhanced by incorporation of an additional secondary amine group on the chain
(A122) (Table 4-1; VIII.104, K = 13.0 pig/ml). We then reasoned that if we increased the
polymer backbone length while maintaining the same percentage of amine along the
backbone, we would detect improved binding because of the greater local positive charge
density. Instead, we found that as backbone length increased, binding of siDNA
decreased (Figure 4-5), perhaps indicating that the longer backbone polymers could not
form the same structures when condensing the siDNA. As a standard for comparison for
the binding affinities we were measuring, we performed the gel shift assay using the
traditional polymers LPEI, BPEI, and poly-L-lysine (PLL) (Table 4-1). Our NPs with the
highest measured binding affinities bound the siDNA nearly as tightly as these control
polymers.

Polyethylene glycol (PEG) is commonly added to many of the polymers used for
nucleic acid delivery because of its supposed benefits for charge shielding (Wolff and

Rozema, 2008). We therefore included PEG on the majority of our NPs. Initially, either

106 A

 

 

 

 

6'0 ' 120
Backbone Length, b

6'1

Figure 4-5 N anoparticle binding affinity relative to polymer backbone length.

NPs were synthesized with increasing backbone length and then tested for siDNA
binding via gel shift assay. Binding affinity (K) was then plotted versus the backbone
length. Better binding (smaller K values) correlated with shorter polymers.

107

 

8 or 115 repeat units of PEG were added directly to one end of the polymer backbone
while keeping the compositions along the backbone constant. For a range of backbone
compositions, PEG caused a reduction in binding affinity (Figure 4-6). This trend was
also present when PEG groups were placed along the polymer backbone (Table 4.1;
compare VIII.65, K = 173 ug/ml, to VIII.28, K = 31.5 ug/ml; and data not shown).

Having established that primary and secondary amines are required for siDNA binding,
and that PEG decreased the overall binding affinity of the NPs, we next looked at how
the remainder of the reactive alkyne sites could be filled to possibly improve binding or
alter function. A series of NPs were synthesized that contained sites filled with a O-, 4-,
10-, or 16-carbon (alkyl) chain such that all sites along the backbone were filled (i.e., the
amine plus alkyl percent filled equals 100%; y + z = 100). Intriguingly, as the alkyl chain
length increased, so too did the binding affinity (Figure 4-7). We hypothesize this result
was due to the longer hydrophobic chains inducing vesicle-like formation around the
siDNA, therefore making the NP better able to condense multiple siDNA molecules per
complex. This was supported by NP size data measured with dynamic light scattering

(DLS) where complex diameter was a linear function of alkyl length for the family of

NPs with eight repeat units of PEG (PEGgPG45(A1230C20) where C was the type and

length of the alkyl chain; Figure 4-8). Size measurements for the other NPs did not yield
clear patterns (data not shown), perhaps because of the disruption of stable structures due
to the longer PEG chains on those NPs. To get a second, independent confirmation of NP
size and shape, we imaged the NPs alone or complexed with siDNA by transmission
electron microscopy (TEM) and estimated the sizes fi'om the images. As observed with

DLS, the particles consistently ranged in size from ~15-150 nm, depending on the

108

particle chemistry (Figure 4-9). However, in all cases the NPs and complexes appeared
to be spherical, as would be expected for vesicles.

Since our NPs were sufficiently interacting with siDNA and the complexes were
on the nanometer scale making cell uptake feasible, we attempted to deliver siRNAs
targeting the Enhanced Green Fluorescent Protein (EGF P) with our NPs into human lung
cancer cells (H1299-EGFP) stably expressing the EGFP reporter plasmid. The siRNA
was fluorescently tagged (red) so we could monitor delivery and localization. Although
the siRNA was observed by fluorescence microscopy to enter the cells (Figure 4-2A), and
the cells appeared healthy (Figure 4-2B), no silencing of the target EGF P was detected at
any of the siRNA or NP concentrations tested (Figure 4—2C). At higher NP
concentrations, the NP/siRNA complexes became toxic (Figure 4-10), with only a few
chemistries leading to higher levels of toxicity than others (data not shown).

Our novel poly(propargyl glycolide) (PPGL) polymer backbone requires primary
and secondary amine containing groups added to the backbone to facilitate siRN A
binding. Indeed, when sufficient positive charge is present, the NPs can bind siRNA (or
the analogous siDNA) at a level approaching what is observed with polymers currently in
use in the literature. The complexes that form are on the nanometer scale where the size
appears to depend on the alkyl chain length. More importantly, NP/siRNA complexes
are readily taken up by the human cells tested here, but unfortunately, we have not yet
observed gene silencing. One possibility is that our NPs are binding siRNA too strongly
and the complexes are in such a conformation that release of the siRN A, either by charge
neutralization or degradation of the polymer, does not occur on the time scale of the

experiments. Further work will be necessary to determine why silencing is not observed

109

so that the polymers can be modified to overcome this problem. Overall, the results
shown here will be useful in designing polymeric drug delivery systems for any nucleic

acid cargo.

110

 

 

 

 

 

 

 

 

FO—C1O_ 20
—v—C16_ 20
1041 LA—C10_ 10
= —O—C10_30
—1:1—x _30
1:1 1:1
1033 GD
2 i
E :
a) 0
3
X 102‘: Q
i A ©
V/
101 .© . . . . . .
0 40 80 120

PEG Length (Repeat Units)

Figure 4-6 N anoparticle binding affinity relative to PEG length.

NPs (PEGaPG45(A122Cy)) containing the indicated alkyl groups (X, C10, or C16; y =
10, 20, or 30%) and amine group (A12; 2 = 100%-y) were synthesized with PEG tails
either 8 or 115 repeat units in length (a= 8 or 115). Binding of siDNA was tested via gel

shift assay and the affinity constant (K) was plotted versus the backbone length.
Nomenclature 18 as described m Table 4-1.

111

 

104':

 

 

 

 

 

El a 115
: O a= 8
1031 D
e g '3
X 102';
: G) g]
I o
101 I I I I0 I
0 5 10 15 20

Alkyl Length, C

Figure 4-7 N anoparticle binding affinity relative to alkyl chain length.

NPs (PEGaPG45(A122 ) were synthesized with amine group (A12, 2 = 80%) and alkyl
chains of varying length (C = 0, 4, 10, or 16, y = 20%) attached to the polymer backbone.
PEG tail length was varied (a = 8 or 115) as in Figure 4-6. Binding affinity (K) for
siDNA was determined by gel shift assay and plotted versus the alkyl chain length (C).

112

 

 

200- 1:1 a= 8
o a=115

 

 

 

 

..
|.
I
u
1
.

O
E]

or

 

0‘] ' I ' I ' l ' I' I ' I ' I ' I ' I
o 2 4 6 8 1o 12 14 16 18
Alkyl Length, C

 

Figure 4-8 Nanoparticle size determined by Dynamic Light Scattering (DLS).
The average hydrodynamic diameter (Hd) of the NP/siDNA complexes

(PEGaPG45(A123oC20) formed as in the gel shift assay were measured by DLS for NPs
of varying alkyl lengths (C = O, 4, 10, or 16).

113

 

 

Figure 4-9 Nanoparticle size and shape determined by Transmission Electron
Microscopy (TEM).

The NP/siDNA complexes were formed as in the gel shift assay and then imaged by
TEM. Sizes of ~20—25 nm were estimated from the images for both (A) VIII.54
(PEGgPG45(A1280X20) and (B) VIII.71 (PEG115PG45(A1230X20). Images are at
200,000x magnification. Scale bars are 100 nm.

114

  

 

. —-— PG45[A128OC1EO3C1220]
‘ + PEGBPG45[A12mC8p82O]
0.4- + PEGBPG45[A128OC1620]

- —v— PEGBPG“[A1280C1020]
0-2' —a— PEGBPG45[A12wC420]

‘ —o— PEGSPG45[A1280XZO]

  

EGFP RFU
O
‘3’

 

   

 

 

101 ....102 .......103
NP Concentration (pg/ml)

Figure 4-10 Nanoparticle toxicity.

The NPs were administered to human H1299 cells stably expressing EGFP for 24 hrs and
then cell density was measured by quantification of EGFP levels. The NPs correspond to
VIII.86 (I), VIII.68 (0), VHI.59 (A), VIII.28 (V), VIII.84 (1:1), and VIH.54 (o) as
described in Table 4-1. RFU: Relative fluorescence units.

115

CHAPTER 5 CONCLUSIONS AND FUTURE WORK

5.1 Conclusions

The work presented here was motivated by the potential for siRNAs to be
developed as therapeutics. Although much is already known about the role of siRNAs in
RNAi, a better understanding of how siRNAs induce the pathway is needed before they
can be of clinical use. To that end, it was shown that the secondary structure of the
mRNA site targeted by siRNA has a significant effect on silencing activity. Sites that
tended to lack structure were favorable targets, especially when the lack of structure was
at either end of the target site. Additionally, siRNA guide strands predicted to have less
structure were similarly preferred. Irnportantly, mRNA target site and siRNA guide
strand secondary structures could be readily predicted by a free energy minimization
algorithm. Therefore, this structure information can be incorporated into current siRN A
selection algorithms with little effort and should help to improve the rate at which active
siRNAs are identified.

The manner in which the guide strand of the siRN A is selected for loading onto
Ag02 is also a critical step that can affect siRNA activity. However, in humans it has
been unknown how that is accomplished. It was demonstrated here for the first time that
human TAR RNA binding protein (TRBP), a component of RISC, can sense siRNA
asymmetry, based most likely on both sequence and the overall and relative
thermodynamic stability of the duplex. Surprisingly, TRBP interacted with ssRNAs in a
fashion consistent with duplex recognition, providing further support for the role of

TRBP in guide strand selection as the asymmetry sensor. Furthermore, it was shown that

116

 

TRBP interactions with siRNAs could be altered through the introduction of sequence
mismatches, DNA substitutions, or 5'-O-methyl groups. Accordingly, manipulation of
the siRNA by either sequence or chemical means offers a viable route for engineering
improved siRNA function.

The creation of highly active siRN As is paramount for eventual therapeutic
applications and the results presented here regarding mRN A target structure effects and
asymmetry sensing by TRBP offer two means for improving that process. However, an
equally critical step is to achieve efficient delivery of the siRNA molecules to the cells of
interest. A novel polymer system generated by Gregory Baker's laboratory via “click”
chemistry was presented here and was used to study a range of variables contributing to
siRNA binding, a key first step necessary for eventually achieving cellular uptake and in
viva silencing; such a systematic study has not been previously reported. It was found
that high affinity binding of the nanoparticles with siRNA required a combination of
primary and secondary amine groups on the same side chains off the polymer backbone,
and that binding could be enhanced by simultaneously including alkyl side chains. Other
changes in the polymer also altered the affinity, such as adding increasingly long PEG
chains, either as a side chain or end-tail. Ultimately, uptake of siRNA into cells grown in
culture could be achieved by the nanoparticles with minimal toxicity. Unfortunately, no
target silencing was detected, indicating that further efforts will be required to optimize

this new platform for firnction of the siRNA.

5.2 Future Work

The work presented here by no means solves all the problems related to the

therapeutic application of siRNAs. Rather, many questions remain to be answered.

117

Below, I outline several topics that I feel are a direct extension of results included in this

dissertation that if investigated, should help to answer some of those questions.

5. 2.1 siRNA selection algorithm

Currently, the best siRNA selection algorithms incorporate numerous parameters
that are both empirical and mechanistically relevant (insofar as the RNAi mechanism is
presently understood). They tend to rely most heavily on rules that support incorporation
of the appropriate guide strand, based on siRNA asymmetry, and account for mRNA
target structure effects (Lu and Mathews, 2007; Shao et al., 2007). The results from Ch.
2 and 3 of this dissertation shed some light on both of these aspects of the RNAi pathway
and therefore an attempt should be made to include them into an siRNA selection
algorithm.

Our work here with TRBP suggests that siRNA asymmetry sensing is a real
phenomenon that can be detected by RISC. However, it is unclear how siRN A
asymmetry is sensed by TRBP and, more importantly, how best to predict the effect. Our
results, and those of (Lu and Mathews, 2007), indicate that a single base-pair or nearest-
neighbor energy parameter might be representative of asymmetry. Accordingly, more
studies are warranted to determine which sequence or energy contributions most
accurately correlate with activity so that they can be incorporated into the selection
algorithm. Furthermore, as additional information is generated using chemical- and
sequence-modified siRNAs with recombinant proteins (discussed below), it too can be
included in the algorithm.

Downstream of siRNA asymmetry, RNA secondary structure effects become

significant, as we and others have found (Ameres et al., 2007; Gredell et al., 2008; Lu

118

and Mathews, 2007; Shao et al., 2007). Although we did not do so, it would be
interesting to see how our structure analyses would possibly differ if siRNA asymmetry
were first taken into account. That is, we should repeat our guide strand and mRN A
target site structure analyses after first classifying an siRNA as either asymmetric or
symmetric (by whatever metric is found to be most significant).

It may also be desirable to repeat our RNA structure calculations using a partition
function approach to determine the most likely structures instead of the minimum fiee
energy structure, which may be present with limited frequency within a cell. While
repeating the analysis, it is recommended that shorter mRNA sequences be used, perhaps
~100 nt up- and down-stream of the target site, to reduce the computational time required
for the predictions, but also because it will likely give more realistic predictions (long-
distance interactions are less likely).

A potential limitation of our approach used here is that the structure classification
scheme is qualitative in nature, depending on classifications that are not at present
quantifiable. Therefore, adapting our structure analysis technique to make the structure
classifications quantitative may be worthwhile and may allow it to be easily coupled with
the asymmetry calculation.

Lastly, there is the definite possibility that siRN A asymmetry and the guide strand
and mRNA target site structure calculations will be insufficient alone to predict siRNA
activity. As such, other variables frequently implemented in the literature should be
considered for inclusion in this algorithm (e. g., GC content, internal stability, etc). The
combination of these parameters could result in a viable algorithm founded primarily

upon mechanistically relevant effects that would be of use for siRN A selection.

119

5. 2.2 Further T RBP characterization

We have shown that TRBP alone is capable of sensing siRNA asymmetry and
that binding appears to require at least some double-stranded, A-form RNA. But where
exactly does TRBP bind an siRNA and how does it do so? The location of binding could
be determined by repeating the UV-crosslinking experiments with siRNAs containing the
4-thiouracil group in different positions along the duplex. I would anticipate that TRBP
most significantly interacts with the ends of the siRNA, with minimal interactions near
the Ag02 cleavage site in the center of the duplex and consistent with what (Tornari et al.,
2004) observed for R2D2. However, based on a model proposed by (Erard et al., 1998), I
speculate that TRBP will hydrogen bond between lysine or arginine residues in TRBP
and guanine nucleotides in the siRNA. Therefore, binding/crosslinking may depend on
stretches of G:C base-pairs or the positioning of G:C pairs relative to each other.

It is also possible that TRBP nonspecifically recognizes the two nucleotide
overhangs on either 3'-end of the siRNAs or the adjacent 5'-terminal nucleotide.
Recognition of overhangs has previously been documented for the PAZ domain of Dicer
(Ma et al., 2004; Ma et al., 2005), and it would not be surprising if TRBP did so as well.
Our results here with the 4-thiouracil crosslinking agent located at position 20 (the first
nucleotide in the overhang) suggest that TRBP is at least in close proximity to the
overhangs and determining if the overhangs are necessary would easily be tested within
the framework of our current experimental design by using siRNAs lacking overhangs.

Given the relative ease with which our binding/crosslinking experiments can be
performed, a more systematic study could be undertaken to test additional siRN A

chemical modifications (as outlined in Ch. 1) to identify their potential uses for dictating

120

 

siRNA asymmetry. Also of interest would be to see if TRBP can asymmetrically
bind/crosslink to miRNAs to determine the role that TRBP plays in that related but albeit
slightly different pathway. Collectively, this work should provide insight into how TRBP

recognizes asymmetry and possibly highlight its broader role as a component of RISC.

5. 2.3 Characterization of additional RNAi-related proteins

Human RISC is known to contain Dicer, TRBP, and Ag02 at a minimum
(MacRae et al., 2008). However, several other proteins, including but not necessarily
limited to, PACT, PKR, and RHA, may also interact directly with the complex and
participate in the RNAi pathway. Many of the proteins contain dsRBDs that should
mediate interactions with siRN As, while others form protein-protein interactions directly
through their respective C-terminal domains (the 'Medipal' domain) (Laraki et al., 2008).
Although much is known about how these individual proteins function, their collective
roles when combined into a multimeric protein complex during RN Ai have not been
studied.

The following experiments that I outline would require that some or all of these
proteins be recombinantly expressed and purified. This task may not be as daunting as it
initially appears. Our laboratory already has expressed and purified TRBP and PKR
(from plasmids obtained fi'om (Daher et al., 2001) and (Bevilacqua and Cech, 1996),
respectively). PACT and RHA are similar to TRBP and PKR (they all contain multiple
dsRBDs), which suggests that they could be relatively easily expressed and purified as
either His- or MBP-tagged proteins, depending on their solubility. Instead, the most
difficult tasks would likely be the purification of Dicer (it needs to be expressed in insect

cells to ensure appropriate levels of post-translational modification) and Ag02 (it would

121

probably contain contaminating ssRNAs). However, both have previously been purified
by (MacRae et al., 2008), suggesting that it is possible to obtain purified Dicer and Ag02.

The first experiments would be to determine if each of the proteins alone can
sense siRNA asymmetry, as TRBP does. If the work proposed in Ch. 5.2.2 is carried out,
this could logically be extended for the other proteins, as well, to clarify where and how
the proteins recognize siRNA (or ssRNA). Based on the similarity of PACT, the
activator of PKR, to TRBP (~42% identical) (Haase et al., 2005), there is a strong
possibility that it too can sense siRN A asymmetry. This result would be particularly
interesting if PACT provides an alternative means for selecting the siRNA guide strand.
Also, since both TRBP and PACT have been shown to interact directly with Dicer, and in
Drosophila, Dicer-2 appears to recognize the end opposite to R2D2 (T omari et al., 2004),
it would be straightforward to verify that a similar effect occurs in humans.

The second set of experiments would be designed to determine the temporal
sequence of initial RLC formation, followed by final active RISC loading. It should be
possible to determine which of the trio of RISC proteins (Dicer/TRBP/AgoZ) first binds
the siRNA, and where, and finally, when the guide strand gets loaded onto Ag02. The
remaining proteins (PACT, RHA, or PKR) could then be added at various points to study
their effects on the process. For example, RHA, as an RNA helicase, may enhance guide
strand loading onto Ag02 by unwinding the siRNA duplex.

Taken together, these experiments would effectively reconstitute RISC and make
it possible to clarify how the siRNA guide strand is initially loaded and subsequently

used to dictate RISC targeting to mRNAs.

122

. ,.,._..__._,.1

5. 2.4 Further nanoparticle investigations

Just as it is important to develop a better understanding of how RISC forms and
guides mRNA target cleavage, considerably more work is also needed if the polymeric
nanoparticles we have developed are to be of value for delivering siRNAs in vitro. Our
initial results show that the NP/siRNA complexes are taken up by the cells but do not
cause appreciable reduction of the target protein levels. Therefore, additional studies are
required to establish methods for improving silencing when initiated via NP-delivered
siRNAs.

One possibility for achieving desirable levels of silencing is to characterize how
the NP/siRNAs enter the cells so that cytoplasmic localization can be maximized.
Polymers have been shown to enter cells through the endosomal pathway where their
cargo can be trapped and prevented access to the cytosol (Douglas et al., 2008;
Yezhelyev et al., 2008). Studies to track uptake typically utilize inhibitors of
endosomes/endosomal acidification so that fluorescently tagged molecules can be
monitored by confocal microscopy. These experiments could easily be repeated with our
fluorescently tagged NPs or siRNAs. For NP/siRNA complexes that are detained in the
endosomes, the polymers can be modified with pH sensitive moieties that can facilitate
endosomal escape, such as was engineered by (Rozema et al., 2003; Rozema et al., 2007;
Wakefield et al., 2005; Xiong et al., 2007), without causing excessive toxicity. After
determining if our NPs are localized in the endosomes, we could begin to modify the
polymer with these kinds of acid-labile groups to potentially enhance their release into

the cytoplasm.

123

It is also possible that our NPs already enter the cytoplasm, and as a result of the
improved siRNA binding we achieved through modifications to the polymer, we may
have simultaneously made it more difficult for the siRNAs to be released. In that case, a
systematic approach to investigating NP degradation/siRN A release is appropriate.
Theoretically, these polymers are degradable into lactic acid derivatives (Gregory Baker,
personal communication). But we do not know on what time scale this might occur. A
useful study would be one that quantifies polymer degradation over time in various
solutions, such as water, buffer, and possibly cell extract. It may be possible to similarly
assess siRN A release from these complexes by performing gel shift assays and then
monitoring for fiee siRNA (siDNA) over time.

As an alternative method for accelerating siRN A release, the siRNA could be
attached directly to the polymer through disulfide bonds that would be subsequently
reduced in the cytoplasm. This method has also been pursued by (Rozema et al., 2007)
and is worth considering. Of course, if successful delivery and silencing can be achieved
with any one of the related NPs derived in this study, a nearly endless list of experiments
can be envisioned, perhaps culminating with an effective in vivo delivery study and

eventually a clinical trial.

124

" "'a'mil

APPENDICES

Materials and Methods for Ch. 2

RNA secondary structure prediction and siRN A selection
The predicted secondary structure of the EGFP mRNA coding sequence was

obtained using default settings on the mfold web server version 3.2

(http://bioinfo.rpi.edu/applications/mfold) or default settings on UNAFold version 3.4, an

 

' r.“ 1."...
.

updated algorithm replacing mfold (Markham and Zuker, 2005; Zuker, 2003). siRNAs [
targeting the EGF P transcript were identified using Ambion Inc.'s siRN A Target Finder I
(http://www.ambion.corn/techlib/misc/siRNA_finder.html). A subset of these siRNAs
was further selected according to the structure of the mRN A target as predicted for the
minimum free energy (MFE) structure (Figures 2-1 and 2-2). Chemically synthesized
siRN As with 3'-UU overhangs were purchased fi'om Dharrnacon. siRN A sequences and
GC content are available in Table 2-1, and the predicted EGF P mRNA structures targeted
by the siRNAs are shown in Figure 2-2. All siRNAs are referred to relative to the target
region, where the 5'-end is position 1, which corresponds to position 19 of the antisense
siRNA strand.
Cell culture and transfection

Human cervical carcinoma (HeLa) and human hepatocellular carcinoma (HepG2)
cells were grown in Dulbecco's modified Eagle medium (DMEM; Invitrogen)

supplemented with 10% (v/v) fetal bovine serum (Biomeda), 100 U/ml penicillin and 100

ug/ml streptomycin (Invitrogen) at 37°C in a 5% C02 humidified incubator. For

transfection, HeLa cells were plated in 6-well plates at 200,000 cells/well and HepG2

125

 

cells were plated in 12-well plates at 400,000 cells/well 24 h before use in medium
containing serum but lacking antibiotic. EGFP plasmid (4.0 ug for HeLa, 1.6 pg for
HepG2, dZEGFP (EGFP variant with an ~2 hr half-life), Clontech) and 30 nM siRNA
were transfected using Lipofectamine 2000 (3.5 and 6.0 ul, respectively, Invitrogen) in
OptiMEM (Invitrogen) according to the manufacturer's protocol. The transfection
solution was replaced after four hours with complete culture medium.

Transfection efficiency was monitored by fluorescence microscopy 24 hrs post-
transfection using fluorescently tagged siRN As (targeting position 126 of the EGFP
transcript or a non-targeting sequence (Block-iT Fluorescent 011 go; Invitrogen)). In all
cases, transfection efficiency was > 85%. Consistent cell viability, independent of
siRNA and/or EGFP plasmid transfection, was confirmed by both microscopy and
CellTiter-Blue Cell Viability Assay (Promega) performed according to the manufacturer's
instructions. Efficiency of EGFP plasmid transfection was similarly monitored by
microscopy. EGFP expression and cell viability were consistent throughout the
experiments.

EGFP protein expression quantification

EGFP expression was quantified from live cells 24 hrs post-transfection. Culture
medium was aspirated, cells were washed twice with phosphate buffered saline (PBS),
and then a volume of PBS equal to the culture medium volume was added. Finally,
fluorescence levels were measured using a SPECTRAmax GEMINI EM plate reader
(Molecular Devices) with excitation at 480 nm, emission at 525 nm, and cutoff filter at
515 nm, as recommended by the manufacturer. Relative fluorescence units (RF U) were

scaled to EGFP transfected cells with no siRNA (100%) and mock transfected cells with

126

no siRNA or plasmid (0%). Fluorescence levels fiom mock transfected cells differed by
less than 2% compared to wells containing only PBS. Experiments were repeated at least
eight times (11 Z 8).

Data and analysis

RNA secondary structure predictions were performed on the genes targeted by
siRNAs in the dataset compiled by (Shabalina et al., 2006) that contained 653 siRNAs.
Prior to the analysis, we removed siRNAs targeting four non-human genes and deleted
five genes that contained over 6,000 nts, as this was the maximum length that could be
analyzed by mfold. Two other genes were removed due to difficulty in identifying the
appropriate target mRNA sequences. The final dataset consisted of 548 siRNAs (533
fi'om (Shabalina et al., 2006) and 15 fiom this work) targeting 42 different genes. As
such, we expect that the pool of sequences used in our analyses is still sufficiently large
and targets enough different mRNAs to be representative of the entire siRN A target
landscape.

Structures targeted by siRNAs were identified fi'om the MFE prediction of the
full-length mRNA sequence, as listed by the National Center for Biotechnology
Information (N CBI) on 12 April 2007, and analyzed as described in Ch. 2.3. The MFE
secondary structure of each siRNA guide strand was also determined using default mfold
settings and analyzed as described. Due to the variability in siRNA overhang
composition utilized throughout the literature (i.e., 3’-UU vs. 3'-deT vs. matching the
target site), only the effects of the 19 nucleotide core on target structure were considered.
All other calculations including unpaired Student's t-test to compare two independent

groups were performed using Microsoft Excel.

127

Materials and Methods for Ch. 3

Protein expression and purification

Plasmids encoding TRBP as a fusion product with maltose binding protein (MBP-
TRBP) and MBP alone were kindly provided by Dr. Anne Gatignol and expressed
essentially as described (Laraki et al., 2008). Briefly, plasmids were transformed into

Rosetta2(DE3) competent cells (Novagen). An overnight culture from a single colony

was diluted 1:50 (v/v) and grown for 3-4 hrs at 37°C, or until A600 = 0.6-1.0, and then

induced with 0.3 mM IPTG. After 2 hrs of expression at 37°C, cells were pelleted (4000
rpm for 10 min), resuspended in Column Buffer (20 mM Tris-HCl, 200 mM NaCl, 1 mM
EDTA), lysed via sonication, and then clarified by high speed centrifugation (15,000 rpm
for 25 min). The supernatant was purified ' with an AKTA FPLC (GE Healthcare).
TRBP-MBP and MBP were eluted from a MBPTrap Column (GE Healthcare) with 5
column volumes of Elution Buffer (20 mM Tris-HCl, 200 mM NaCl, 1 mM EDTA, 10
mM Maltose) and stored in Elution Buffer at ~80°C until use.

Protein concentration was measured by Bradford assay (BioRad) or Bicinchoninic
acid (BCA) assay (Pierce; Thermo Scientific). Products were assessed by separating
approximately equal amounts of protein per lane on denaturing linear 4-20% Tris-HCl
gels. Total protein was visualized by staining with Gel Code Blue (Pierce; Thermo
Scientific) and specific proteins were detected via western blot. For western blot, the
separated proteins were transferred to nitrocellulose membranes and then incubated
overnight at 4°C with the appropriate antibodies (MBP (New England Biolabs), TRBP
(Abnova)). Blots were washed with TBS-Tween, incubated for an additional 1 hr with

HRP-linked secondary antibody (Pierce; Thermo Scientific), and developed using the

128

 

SuperSignal West Fernto Maximum Sensitivity Substrate kit (Pierce; Thermo Scientific).
Images were obtained on a BioRad ChemiDoc XRS imager using Quantity One software.
Nucleic acids

DNA oligonucleotides were purchased fi'om Integrated DNA Technologies.
Chemically synthesized RNAs were purchased from Dharmacon as either ready-to—use
siRNAs or ssRNAs. The lists of sequences used are provided in Table 6-1 to 6-5. HIV-l
TAR RNA was prepared by in vitro transcription fi'orn a plasmid kindly provided by Dr.
K-T Jeang. Briefly, 1 pg of linearized plasmid DNA was transcribed with the
Megashortscript Kit (Ambion) following the manufacturer’s instructions. Transcripts
were resolved in 8M urea-10% polyacrylamide gels, visualized by UV shadowing, and
eluted from gel slices by crushing and soaking in TE buffer ( 10 mM Tris, 1 mM EDTA)
for 10 min at 75°C. RNA was then concentrated by ethanol precipitation, washed with
70% ethanol, and quantitated spectrophotometrically (N anoDrop). In order to remove the
5'-triphosphate, RNAs were treated with calf intestinal phosphatase (New England
Biolabs). Treated RNAs were purified first by phenol/chloroform extraction, then

ethanol precipitated, gel purified (as above), and concentrated. The concentration of the
product containing 5'-OH was again measured prior to end-labeling with [7-32P]-ATP

(see below).

129

Table 6-1 Nucleic acids used in Ch. 3 (listed 5' to 3').

 

Strand Name Sequence
pp-luc-ss CGUACGCGGAAUACUUCGAUA
pp-luc-ss-4SU(20) CGUACGCGGAAUACUUCGAUA
pp-luc-as UCGAAGUAUUCCGCGUACGUG

pp-luc-as-4SU(20) UCGAAGUAUUCCGCGUACGUG
pp-luc-ss-A(1) §GUACGCGGAAUACUUCGAUA
pp-luc-ss-C(19) CGUACGCGGAAUACUUCGQUA

sodl -as GUCACAUUGCCCAAGUCUCUU
sodl -as-4SU(20) GUCACAUUGCCCAAGUCUCUU
sodl -ss GAGACUUGGGCAAUGUGACUU

sodl-ss-4SU(20) GAGACUUGGGCAAUGUGACUU
sodl-as-U(l) QAGACUUGGGCAAUGUGACUU
sodl -as-A(19) GAGACUUGGGCAAUGUGALXUU

EGFP-as UGCGCUCCUGGACGUAGCCUU
EGFP-as-4SU(20) UGCGCUCCUGGACGUAGCCUU
EGFP-SS GGCUACGUCCAGGAGCGCAUU

EGFP-ss-4$U(20) GGCUACGUCCAGGAGCGCAUU
EGFP-SS-U(l ) EGCUACGUCCAGGAGCGCAUU
EGFP-ss-C(l9) GGCUACGUCCAGGAGCGCQUU

EGFP-aS-S'OCH3 mTGCGCUCCUGGACGUAGCCUU

EGFP-SS-S'OCH3 mTGCUACGUCCAGGAGCGCAUU

U denotes 4-thiouracil; mT denotes 5'-O-methyl thymidine; underline denotes nucleotide
that is mismatched to its complementary sequence.

130

Table 6-2 Nucleic acid duplexes used in Ch. 3 (top strand listed 5' to 3', bottom

strand listed 3' to 5').

Strand Name

Sequence

 

RN Ass
RNAas

p*-GGCUACGUCCAGGAGCGCAUU
UUCCGAUGCAGGUCCUCGCGU-p*

 

DNAss
DNAas

p*-GGCTACGTCCAGGAGCGCATT
TTCCGATGCAGGTCCTCGCGT-p*

 

RNAss
DNAas

p*-GGCUACGUCCAGGAGCGCAUU
TTCCGATGCAGGTCCTCGCGT-p*

 

DNAss
RNAas

p*-GGCTACGTCCAGGAGCGCATT
UUCCGAUGCAGGUCCUCGCGU-p*

* denotes 5 '-phosphorylation with y-BZP.

131

 

_ l

 

Table 6-3 Nucleic acid duplexes targeting pp-luciferase used in Ch. 3 (top strand

listed 5' to 3', bottom strand listed 3' to 5').

Duplex Name

Strand Name

Sequence

 

a

pp-luc-as
pp-luc-ss

p*-UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGC-p*

 

b

pp—luc-as-4SU(20)
pp-luc-ss

p*-UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGC-p

 

pp-luc-as-4SU(20)
pp-luc-ss

p -UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGC-p*

 

pp-luc-as
pp-luc-ss-4SU(20)

p -UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGC-p*

 

pp-luc-as
pp-luc-ss-4SU(20)

p*-UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGC-p

 

pp-luc-as—48U(20)
pp-luc-ss—4$U(20)

p*-UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGC-p*

 

pp-luc-as-48U(20)
pp-luc-ss-A(1 )

p*-UCGAAGUAUUCCGCGUACGUG
AUAGCUUCAUAAGGCGCAUGA-P

 

pp-luc-as—4SU(20)
pp—luc-ss-C(19)

p*-UCGAAGUAUUCCGCGUACGUG
AUQGCUUCAUAAGGCGCAUGC-P

 

pp-luc-as-4SU(20)
pp-luc-ss

p*-UCGAAGUAUUCCGCGUACGUG
ATAGCTTCATAAGGCGCATGCfip

 

pp-Iuc-as
pp—luc-ss-4$U(20)

p -TCGAAGTATTCCGCGTACGTG
AUAGCUUCAUAAGGCGCAUGC-p*

PT“ i... 2;; M. i

U denotes 4-thiouracil; underline denotes nucleotide that is mismatched to its
complementary sequence; italics denote DNA strands; p denotes 5'-phosphorylation; *

denotes 5'-phosphorylation with y-32P.

132

Table 6—4 Nucleic acid duplexes targeting sodl used in Ch. 3 (top strand listed 5' to

3', bottom strand listed 3' to 5').

Duplex Name

Strand Name

Sequence

 

a

sodl-ss
sodl-as

p*-GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P*

 

b

sodl-ss-4SU(20)
sodl-as

p*-GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P

 

sodl-ss-4SU(20)
sodl-as

p -GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P*

 

sodl-ss
sodl-as-4SU(20)

p -GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P*

 

sodl-ss
sodl-as-4SU(20)

p*-GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P*

 

sodl-ss-4SU(20)
sodl-as-4SU(20)

p*-GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P*

 

sodl-ss-4SU(20)
sodl -as-U(1)

p*-GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUU-P

 

sodl-ss-4SU(20)
sodl-as-A(19)

p*-GAGACUUGGGCAAUGUGACUU
UU§UCUGAACCCGUUACACUG-P

 

sodl-ss-4SU(20)
sodl-as

p*-GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P

 

sodl-s3
sodl-as-4SU(20)

p -GAGACUUGGGCAAUGUGACUU
UUCUCUGAACCCGUUACACUG-P*

U denotes 4-thiouracil; underline denotes nucleotide that is mismatched to its
compleyrrzrentary sequence; p denotes 5'-phosphorylation; * denotes 5'-phosphorylation
with 'y- P.

133

Table 6-5 Nucleic acid duplexes targeting EGFP used in Ch. 3 (top strand listed 5' to

3', bottom strand listed 3' to 5').

Duplex Name

Strand Name

Sequence

 

a

EGFP-as
EGFP-ss

p*-UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p*

 

b

EGFP-as-4SU(20)
EGFP-ss

p*-UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p

 

EGFP-as-4SU(20)
EGFP-ss

p -UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p*

 

EGFP-as
EGFP-ss—4$U(20)

p -UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p*

 

EGFP-as
EGFP-ss-4SU(20)

p*-UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p

 

EGFP-as-4SU(20)
EGFP-ss-4SU(20)

P*-UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p*

 

EGFP-as—4SU(20)
EGFP-ss-U(l )

P*-UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGQ-p

 

EGFP-as-4SU(20)
EGFP-ss-C(19)

p*-UGCGCUCCUGGACGUAGCCUU
UUQCGCGAGGACCUGCAUCGG-p

 

EGFP-as-48U(20)
EGFP-ss-S'OCH3

p*-UGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGTm

 

EGFP-as-S'OCH3
EGFP-ss-4SU(20)

mTGCGCUCCUGGACGUAGCCUU
UUACGCGAGGACCUGCAUCGG-p*

V“ .: “rm".‘Qh-l
4‘ ‘

U denotes 4-thiouracil; mT denotes 5'-O-methyl thymidine; underline denotes nucleotide
that is mismatched to its complementary sequence; italics denote DNA strands; p denotes

5'-phosphorylation; * denotes 5'-phosphorylation with y-3ZP.

134

 

siRN A duplex thermodynamic calculation

The method for calculating the relative thermodynamic asymmetry for siRN A
duplexes was performed as described by (Hutvagner, 2005) using the most recent
nearest-neighbor values available (Mathews et al., 1999). Basically, the fiee energy (AG)
was calculated for each end of the duplex by smnming the four terminal nearest-neighbor
values (stacking energies), the value for the first nucleotide of the overhang (single base
stacking energies), and a penalty of +0.5 kcal/mol for a terminal A:U pairing
(miscellaneous energies). The relative asymmetry (AAG) was then calculated by taking
the difference between the free energy of the AS 5'-end and the flue energy of the SS 5'-
end. For example, for the asymmetric pp-luc siRNA, the AS 5'-end AG = (-2.40) + (-
2.40) + (-2.40) + (-0.90) + (-O.60) + (0.5) = -8.2 kcal/mol, the SS 5'-end AG = (-2.40) + (-
2.20) + (-1.30) + (-2.20) + (-0.6) + (0.0) = -8.7 kcal/mol, yielding a AAG = (-8.2) — (-8.7)
= 0.5 kcal/mol. Since AAG > 0, it indicates that the SS 5'-end is more stably hybridized
than the AS 5'-end, thereby favoring incorporation of the AS into RISC.
Duplex preparation and end-labeling

For Figure 3-2, oligonucleotides were hybridized by mixing equal amounts of
both top and bottom strands in 1x STE buffer (10 mM Tris, 100 mM NaCl, 1 mM EDTA)
and then heated to 90°C for 3 min, followed by incubation at 37°C for 60 min. Products
were verified on native TBE gels stained with SYBR Gold (Invitrogen). Duplexes for

Figure 3-2 and ssRNAs (Figure 3-11) (3 pmol) were directly 5'-radiolabeled with 10

pmol of [y-32P]-ATP using T4 polynucleotide kinase (New England Biolabs);

unincorporated label was removed using G-25 Sephadex columns (Roche Applied

Science). TAR RNA (3 pmol) was similarly labeled following CIP treatment.

135

 

 

For the remaining figures in Ch. 3 where strands were alternately hot and cold
labeled, individual single-strands (3 pmol) were labeled as described above (10 pmol of
ATP) in 25 ul reactions, where nonisotopic ATP was used for cold labeling. The 5'-O-
methyl modified strands were not end-labeled. The two strands comprising the siRN A
duplex were then mixed (50 pl), 5.56 1.11 of 10x STE was added for 1.0x STE final
concentration, and samples were heated to 90°C for 3 min and then 37°C for 60 min.
Unincorporated label was then removed fiom the annealed siRNA using the Sephadex
columns. Final products were verified on native TBE gels.

Native gel shift assay
TRBP binding was determined by gel shift assays using radiolabeled

oligonucleotides present at limiting concentrations relative to protein concentrations.
Binding reactions (10 ul) were prepared with ~1-3x104 cpm of nucleic acid and 350 nM
TRBP (unless noted otherwise) in Elution Buffer supplemented with 20 mM HEPES, 40

mM KCl, 1.5 mM MgC12, 0.1% Nonidet P40, and 10 U SUPERase-In (Ambion).

Samples were incubated at room temperature for 30 min, mixed with 2 ul of 5x Nucleic
Acid Sample Loading Buffer (BioRad), and 10 ul were loaded onto a pre-run native TBE
gel. Mini-gels were prepared where the top half of the gel was 4% (37:1
acrylamidezbisacrylamide) and the bottom half was a gradient between 4-20%. Gels
were run at 150 V for ~40 min, dried on filter paper under vacuum at 80°C for 50 min,
exposed to a storage phosphor screen overnight (~12-16 hrs), and then imaged on a
Storm 860 imager (GE Healthcare). Relative signal intensities were quantified using

Quantity One software and normalized within a single lane. Native gels for loading

136

r

control were similarly prepared but with the volume of TRBP replaced with an equal
volume of Elution Buffer.
UV crosslinking, denaturing gel shift assay

TRBP crosslinking to nucleic acids was measured as in the native gel shift assay
except that after the initial 30 min binding, the samples were subsequently exposed to
312nm UV light (Fisher Scientific Electrophoresis Systems Transilluminator) for 10 min
while on an ice cold aluminum block (4°C). In order to minimize non-specific
crosslinking, the samples were covered by a Petri dish to block shorter wavelengths.
Samples were then mixed with 5 ul of 3x SDS Sample Buffer (New England Biolabs),
boiled at 95°C for 3 min, collected by brief centrifugation, and 12 pl were loaded onto a
pre-run non-linear denaturing 4-20% Tris-HCl gel. Gels were run, dried, imaged, and
analyzed as described above. Statistical analyses are reported as p-values fiom unpaired,

two-tailed Student's t-test performed using Microsoft Excel.

Materials and Methods for Ch. 4

Nucleic acids

Three types of nucleic acids were considered: deoxyribonucleic acids (DNA),
plasmid DNA (pDNA), and short interfering ribonucleic acids (siRNA). DNA was
chemically synthesized by Integrated DNA Technologies (IDT) as 21 nt long sequences
(5'-CTGGGTGCTCAGGTAGTGGTT-3', sense strand; 5'-
CCACT ACCT GAGCACCCAGTT-3', antisense strand) that were complementary. When
hybridized, the dsDNA product was 21 base-pairs long with TT overhangs on both 3'-

ends, similar to the structure characteristic of siRNAs. When noted, the DNA was

137

:1

modified with a 6-FAM on the 5'-end so that is would provide green fluorescence when
excited by UV light. Plasmid DNA encoding the destabilized version of the Enhanced
Green Fluorescent Protein (EGFP; d2EGFP, Clontech) such that the protein half-life was
~2hrs was prepared following standard protocols. siRNAs targeting the EGFP were
chemically synthesized by Dharmacon, Inc. and the sequences have been published
previously (Gredell et al., 2008). A non-targeting siRNA containing a F ITC modification
on both 5'-ends for green fluorescence imaging was purchased fi'om Invitrogen (Block-iT
Fluorescent Oligo).
Polymeric nanoparticle synthesis

Nanoparticles (NPs) were synthesized following the methods described in (J iang
et al., 2008) and in more detail elsewhere (Gregory Baker, personal communication).
Gel shift assay

An siDNA 21 nt in length and possessing 2 nt overhangs on either 3'-end
(siDNA), analogous to an siRNA, was prepared by annealing complementary strands
according to standard techniques described in Appendices — Materials and Methods for
Ch. 3. To simplify imaging, both strands of the siDNA contained 5'-6-FAM
modifications.

NP or control polymer binding of siDNA was measured using a gel shift assay.
Briefly, 25 pl reactions were prepared in water with 200 nM dsDNA and NP

concentrations typically ranging from 0.10-7500 pg/mL. Reactions were incubated at

room temperature for 15 min before addition of 6.25 pl of 5x Nucleic Acid Loading.

Buffer (BioRad). An aliquot of each sample (15 pl) was loaded onto a 0.8% agarose gel

(prepared in 45 mM Tris-Borate, 1 mM EDTA buffer) and electrophoresed at 50 V for

138

 

~35 min, or until the first blue front was ~2/3 down the gel. Gels were imaged and
analyzed on a BioRad Cherrridoc XRS machine. Band intensities for unbound siDNA
from lanes with NP were normalized to lanes containing only siDNA to determine the
fiaction of unbound siDNA (fraction unbound = [intensity of unbound siDNA from NP
treated lane]/ [intensity of untreated siDNA]). The fraction of bound siDNA was
calculated fiom the unbound fraction (fraction bound siDNA = 1 — fraction unbound).
Binding constants (K) were determined for each NP by fitting the fraction of siDNA
bound versus NP concentration to the Hill equation using Origin (Microcal Software,
Inc.).
Dynamic light scattering and transmission electron microscopy

NP/siDNA complex sizes were measured using dynamic light scattering (DLS)
and transmission electron microscopy (T EM) by first forming complexes as in the gel
shift assay and then following standard methods to be described elsewhere that were
specific to the particular technique (Gregory Baker, personal communication).
Application of nanoparticle/siRN A complexes to cell culture

The cellular response to NP or NP/siRNA complexes was evaluated by treating
the NP as if it were a lipid-based transfection and then applying it to cells grown in
culture according to standard transfection techniques (see Appendices — Materials and
Methods for Ch. 2 or www.invitrogen.com for details).

Initial validation of rhodamine-tagged NP uptake was performed after incubating
HeLa (human cervical carcinoma) cells with NP rrrixtures up to ~1.0 mg/ml for 24 hrs.

Cells were washed to remove NPs remaining in solution and then imaged by fluorescence

139

 

s“
i
‘l
i
1
4'.

 

microscopy to monitor the localization and intensity of rhodamine (red) signal (Figure 4-
2A).

The NP chemistries that were capable of binding siDNA were tested for their
effectiveness to facilitate siRN A uptake into H1299 cells (kindly provided by Prof. Claus
Bus) stably expressing an EGFP plasmid related to the one described in Appendices —
Materials and Methods for Ch. 2. A Dy547-modified siRNA synthesized by Dharmacon
(30 nM) and known to effectively silence the target EGFP mRNA was mixed with ~1.0-
5,000 pg/ml NP as in the gel shift assay using filter sterilized (< 0.22 pm) solutions
whenever possible. Total volumes were scaled according to the volume of medium in
which the cells were cultured to achieve the desired final concentrations. NP/siRN A
complexes were applied to the H1299 cells for 24 hrs and then cells were imaged by
fluorescence microscopy to assess EGFP knockdown (loss of green signal) and siRNA
localization (red signal from Dy547). To measure NP or NP/siRNA toxicity profiles, the
EGFP signal from the H1299 cells was used as an indicator of total cell number. H1299
cells grown in 96-well plates were treated in triplicate with NP or NP complexed with a
non-targeting siRNA and EGFP levels were quantified fiom live cells as in Appendices —
Materials and Methods for Ch. 2. EGFP relative fluorescence units (RFU) were

calculated by normalizing levels for treated samples to untreated control samples.

140

 

BIBLIOGRAPHY

Akinc, A., Zumbuehl, A., Goldberg, M., Leshchiner, E.S., Busini, V., Hossain, N.,
Bacallado, S.A., Nguyen, D.N., Fuller, J., Alvarez, R., et al. (2008). A combinatorial
library of lipid-like materials for delivery of RNAi therapeutics. Nat Biotechnol 26, 561-
569.

Allen, R.S., Miller, J .A., Chitty, J .A., Fist, A.J., Gerlach, W.L., and Larkin, P.J. (2008).
Metabolic engineering of morphinan alkaloids by over-expression and RN Ai suppression
of salutaridinol 7-O-acetyltransferase in opium poppy. Plant Biotechnol J 6, 22-30.

Allen, R.S., Millgate, A.G., Chitty, J.A., Thisleton, J., Miller, J.A., Fist, A.J., Gerlach,
W.L., and Larkin, P.J. (2004). RNAi-mediated replacement of morphine with the
nonnarcotic alkaloid reticuline in opium poppy. Nat Biotechnol 22, 1559-1566.

Amarzguioui, M., and Prydz, H. (2004). An algorithm for selection of functional siRNA
sequences. Biochem Biophys Res Commun 316, 1050-1058.

Ameres, S.L., Martinez, J., and Schroeder, R. (2007). Molecular basis for target RNA
recognition and cleavage by human RISC. Cell 130, 101-112.

Bartlett, D.W., and Davis, ME. (2007a). Effect of siRNA nuclease stability on the in
vitro and in vivo kinetics of siRNA-mediated gene silencing. Biotechnol Bioeng 97, 909-
921.

Bartlett, D.W., and Davis, ME. (2007b). Physicochemical and biological characterization
of targeted, nucleic acid-containing nanoparticles. Bioconjug Chem 18, 456-468.

Baulcombe, D. (2004). RNA silencing in plants. Nature 431, 356-363.

Benkirane, M., Neuveut, C., Chun, R.F., Smith, S.M., Samuel, C.E., Gatignol, A., and
Jeang, K.T. (1997). Oncogenic potential of TAR RNA binding protein TRBP and its
regulatory interaction with RN A-dependent protein kinase PKR. EMBO J I 6, 611-624.

Berkhout, B., and J eang, K.T. (1991). Detailed mutational analysis of TAR RNA: critical
spacing between the bulge and loop recognition domains. Nucleic Acids Res 19, 6169-
6176.

Bernstein, E., Caudy, A.A., Hammond, S.M., and Harmon, G]. (2001). Role for a
bidentate ribonuclease in the initiation step of RNA interference. Nature 409, 363-366.

Bevilacqua, P.C., and Cech, TR. (1996). Minor-groove recognition of double-stranded

RNA by the double-stranded RNA-binding domain from the RNA-activated protein
kinase PKR. Biochem 35, 9983-9994.

141

h

Bevilacqua, P.C., George, C.X., Samuel, CE, and Cech, TR. (1998). Binding of the
protein kinase PKR to RN As with secondary structure defects: role of the tandem A-G
mismatch and noncontiguous helixes. Biochem 3 7, 6303-6316.

Bohula, E.A., Salisbury, A.J., Sohail, M., Playford, M.P., Riedernann, J., Southern, E.M.,
and Macaulay, V.M. (2003). The Efficacy of Small Interfering RNAs Targeted to the
Type 1 Insulin-like Growth Factor Receptor (IGFlR) Is Influenced by Secondary
Structure in the IGFlR Transcript. J Biol Chem 278, 15991-15997.

Bridge, A.J., Pebemard, S., Ducraux, A., Nicoulaz, A.-L., and Iggo, R. (2003). Induction
of an interferon response by RNAi vectors in mammalian cells. Nat Genet 34, 263-264.

Brown, K.M., Chu, C.Y., and Rana, TM. (2005). Target accessibility dictates the
potency of human RISC. Nat Struct Mol Biol 12, 469-470.

Brus, C., Petersen, H., Aigner, A., Czubayko, F ., and Kissel, T. (2004). Physicochemical
and Biological Characterization of Polyethylenimine-grafi-Poly(ethylene glycol) Block
Copolymers as a Delivery System for Oligonucleotides and Ribozymes. Bioconjug Chem
15, 677-684.

Carlson, C.B., Stephens, O.M., and Beal, RA. (2003). Recognition of double-stranded
RNA by proteins and small molecules. Biopolymers 70, 86-102.

Candy, A.A., and Hannon, G]. (2004). Induction and biochemical purification of RNA-
induced silencing complex fi'om Drosophila 82 cells. Methods Mol Biol 265, 59-72.

Caudy, A.A., Ketting, R.F., Hammond, S.M., Denli, A.M., Bathoom, A.M., Tops, B.B.,
Silva, J .M., Myers, M.M., Harmon, G.J., and Plasterk, RH. (2003). A micrococcal
nuclease homologue in RNAi effector complexes. Nature 425, 411-414.

Caudy, A.A., Myers, M., Harmon, G.J., and Hammond, SM. (2002). Fragile X-related
protein and VIG associate with the RNA interference machinery. Genes Dev 16, 2491-
2496.

Chang, K.-Y., and Ramos, A. (2005). The double-stranded RNA-binding motif, a
versatile macromolecular docking platform. FEBS J 272, 2109-2117.

Chen, P.Y., Weinmann, L., Gaidatzis, D., Pei, Y., Zavolan, M., Tuschl, T., and Meister,
G. (2008). Strand-specific 5'-O-methylation of siRNA duplexes controls guide strand
selection and targeting specificity. RNA 14, 263-274.

Chendrimada, T.P., Gregory, R.I., Kurnaraswamy, E., Norman, J ., Cooch, N., Nishikura,
K., and Shiekhattar, R. (2005). TRBP recruits the Dicer complex to Ag02 for microRNA
processing and gene silencing. Nature 436, 740-744.

Chiu, Y., and Rana, TM. (2003). siRNA function in RNAi: A chemical modification
analysis. RNA 9, 1034-1048.

142

 

Clamme, J .P., Azoulay, J ., and Mely, Y. (2003). Monitoring of the formation and
dissociation of polyethylenimine/DNA complexes by two photon fluorescence correlation
spectroscopy. Biophys J 84, 1960-1968.

Conklin, D.S. (2003). RNA-Interference-Based Silencing of Mammalian Gene
Expression. ChemBioChem 4, 1033-1039.

Consortium (2004). Finishing the euchromatic sequence of the human genome. Nature
431, 931-945.

Corey, DR. (2007). Chemical modification: the key to clinical application of RNA
interference? J Clin Invest 11 7, 3615-3622.

Cosentino, G.P., Venkatesan, S., Serluca, F.C., Green, S.R., Mathews, M.B., and
Sonenberg, N. (1995). Double-stranded-RNA-dependent protein kinase and TAR RNA-
binding protein form homo- and heterodirners in vivo. Proc Natl Acad Sci U S A 92,
9445-9449.

Cox, K.M., Sterling, J .D., Regan, J .T., Gasdaska, J .R., Frantz, K.K., Peele, C.G., Black,
A., Passmore, D., Moldovan-Loomis, C., Srinivasan, M., et al. (2006). Glycan
optimization of a human monoclonal antibody in the aquatic plant Lemna minor. Nat
Biotechnol 24, 1591-1597.

Daher, A., Longuet, M., Dorin, D., Bois, F., Segeral, E., Bannwarth, S., Battisti, P.L.,
Purcell, D.F., Benarous, R., Vaquero, C., et al. (2001). Two dimerization domains in the
trans-activation response RNA-binding protein (TRBP) individually reverse the protein
kinase R inhibition of HIV-1 long terminal repeat expression. J Biol Chem 276, 33899-
33905.

Daviet, L., Erard, M., Dorin, D., Duarte, M., Vaquero, C., and Gatignol, A. (2000).
Analysis of a binding difference between the two dsRNA-binding domains in TRBP
reveals the modular function of a KR-helix motif. Eur J Biochem 267, 2419-2431.

de Fougerolles, A., Vomlocher, H.P., Maraganore, J ., and Lieberman, J. (2007).
Interfering with disease: a progress report on siRNA-based therapeutics. Nat Rev Drug
Discov 6, 443-453.

de Fougerolles, AR. (2008). Delivery vehicles for small interfering RNA in vivo. Hum
Gene Ther 19, 125-132.

de Martimprey, H., Vauthier, C., Malvy, C., and Couvreur, P. (2009). Polymer
nanocarriers for the delivery of small fragments of nucleic acids: oligonucleotides and
siRNA. Eur J Pharrn Biopharrn 71, 490-504.

Ding, Y., Chan, C.Y., and Lawrence, CE. (2004). Sfold web server for statistical folding
and rational design of nucleic acids. Nucleic Acids Res 32, W135-141.

143

Dodo, H.W., Konan, K.N., Chen, F.C., Egnin, M., and Viquez, OM. (2008). Alleviating
peanut allergy using genetic engineering: the silencing of the irnmunodominant allergen
Ara h 2 leads to its significant reduction and a decrease in peanut allergenicity. Plant
Biotechnol J 6, 135-145.

Doench, J.G., and Sharp, RA. (2004). Specificity of microRNA target selection in
translational repression. Genes Dev 18, 504-511.

Donker, R.B., Mouillet, J .-F., Nelson, D.M., and Sadovsky, Y. (2007). The expression of
Argonaute2 and related microRNA biogenesis proteins in normal and hypoxic
trophoblasts. Mol Hum Reprod 13, 273-279.

Doshi, K., Cannone, J ., Cobaugh, C., and Gutell, R. (2004). Evaluation of the suitability
of flee-energy minimization using nearest-neighbor energy parameters for RNA
secondary structure prediction. BMC Bioinform 5, 105.

Douglas, K.L., Piccirillo, C.A., and Tabrizian, M. (2008). Cell line-dependent
internalization pathways and intracellular trafficking determine transfection efficiency of
nanoparticle vectors. Eur J Pharrn Biopharm 68, 676-687.

Doyle, M., and Jantsch, M.F. (2002). New and old roles of the double-stranded RNA-
binding domain. J Struct Biol 140, 147-153.

Eckrnann, CR, and Jantsch, M.F. (1997). Xlrbpa, a Double-stranded RNA-binding
Protein Associated with Ribosomes and Heterogeneous Nuclear RNPs. J Cell Biol 138,
239-253.

Elbashir, S.M., Harborth, J ., Weber, K., and Tuschl, T. (2002). Analysis of gene function
in somatic mammalian cells using small interfering RNAs. Methods 26, 199-213.

Elbashir, S.M., Lendeckel, W., and Tuschl, T. (2001a). RNA interference is mediated by
21- and 22-nucleotide RNAs. Genes Dev 15, 188-200.

Elbashir, S.M., Martinez, J., Patkaniowska, A., Lendeckel, W., and Tuschl, T. (2001b).
Functional anatomy of siRNAs for mediating efficient RNAi in Drosophila melanogaster
embryo lysate. EMBO J 20, 6877-6888.

Elrnen, J ., Thonberg, H., Ljungberg, K., Frieden, M., Westergaard, M., Xu, Y., Wahren,
B., Liang, Z., Orum, H., Koch, T., et al. (2005). Locked nucleic acid (LNA) mediated
improvements in siRNA stability and functionality. Nucleic Acids Res 33, 439-447.

Erard, M., Barker, D.G., Amalric, F., Jeang, K.-T., and Gatignol, A. (1998). An Arg/Lys-

rich core peptide mimics TRBP binding to the HIV-1 TAR RNA upper-stern/loop. J Mol
Biol 279, 1085-1099.

144

 

 

Far, R., and Sczakiel, G. (2003). The activity of siRNA in mammalian cells is related to
structural target accessibility. a comparison with antisense oligonucleotides. Nucleic
Acids Res 31, 4417-4424.

Fedorov, Y., Anderson, E.M., Birmingham, A., Reynolds, A., Karpilow, J ., Robinson, K.,
Leake, D., Marshall, W.S., and Khvorova, A. (2006). Off-target effects by siRNA can
induce toxic phenotype. RNA 12, 1188-1 196.

Felgner, P.L., Gadek, T.R., Holrn, M., Roman, R., Chan, H.W., Wenz, M., Northrop, J .P.,
Ringold, G.M., and Danielsen, M. (1987). Lipofection: a highly efficient, lipid-mediated
DNA-transfection procedure. Proc Natl Acad Sci U S A 84, 7413-7417.

Fierro-Monti, I., and Mathews, MB. (2000). Proteins binding to duplexed RNA: one
motif, multiple firnctions. Trends Biochem Sci 25, 241-246.

Fire, A., Xu, S., Montgomery, M.K., Kostas, S.A., Driver, SE, and Mello, CC. (1998).
Potent and specific genetic interference by double-stranded RNA in Caenorhabditis
elegans. Nature 391, 806-811.

Forstemann, K., Tomari, Y., Du, T., Vagin, V.V., Denli, A.M., Bratu, D.P., Klattenhoff,
C., Theurkauf, W.E., and Zamore, PD. (2005). Normal microRNA maturation and germ-
line stem cell maintenance requires Loquacious, a double-stranded RNA-binding domain
protein. PLoS Biol 3, e236.

Garcia, M.A., Gil, J ., Ventoso, I., Guerra, S., Domingo, E., Rivas, C., and Esteban, M.
(2006). Impact of Protein Kinase PKR in Cell Biology: fiom Antiviral to
Antiproliferative Action. Microbiol Mol Biol Rev 70, 1032-1060.

Gatignol, A., Buckler-White, A., Berkhout, B., and Jeang, K.T. (1991). Characterization
of a human TAR RNA-binding protein that activates the HIV -1 LTR. Science 251, 1597-
1600.

Gatigrrol, A., Buckler, C., and Jeang, K.T. (1993). Relatedness of an RNA-binding motif
in human immunodeficiency virus type 1 TAR RN A-binding protein TRBP to human
Pl/dsI kinase and Drosophila staufen. Mol Cell Biol 13, 2193-2202.

Ge, G., Wong, G.W., and Luo, B. (2005). Prediction of siRNA knockdown efficiency
using artificial neural network models. Biochem Biophys Res Commun 336, 723-728.

Geng, C.M., and Ding, H.L. (2008). Double-mismatched siRNAs enhance selective gene
silencing of a mutant ALS-causing allele. Acta pharrnacologica Sinica 29, 211-216.

Goutelle, S., Maurin, M., Rougier, F ., Barbaut, X., Bourguigrron, L., Ducher, M., and

Maire, P. (2008). The Hill equation: a review of its capabilities in pharmacological
modelling. Fundamental & clinical pharmacology 22, 633-648.

145

Gredell, J .A., Berger, A.K., and Walton, SP. (2008). Impact of target mRNA structure on
siRNA silencing efficiency: A large-scale study. Biotechnol Bioeng 100, 744—755.

Gregory, R.I., Chendrimada, T.P., Cooch, N., and Shiekhattar, R. (2005). Human RISC
Couples MicroRNA Biogenesis and Posttranscriptional Gene Silencing. Cell 123, 631-
640.

Grimm, D., Streetz, K.L., Jopling, C.L., Storm, T.A., Pandey, K., Davis, C.R., Marion,
P., Salazar, F., and Kay, M.A. (2006). Fatality in mice due to oversaturation of cellular
microRNA/short hairpin RNA pathways. Nature 441, 537-541.

Haase, A.D., J askiewicz, L., Zhang, H., Laine, S., Sack, R., Gatignol, A., and Filipowicz,
W. (2005). TRBP, a regulator of cellular PKR and HIV -1 virus expression, interacts with
Dicer and functions in RNA silencing. EMBO Rep 6, 961-967.

Haley, B., and Zamore, PD. (2004). Kinetic analysis of the RNAi enzyme complex. Nat
Struct Mol Biol 11, 599-606.

Harmon, GJ. (2002). RNA interference. Nature 418, 244-251.

Hargittai, M.R., Gorelick, R.J., Rouzina, I., and Musier-Forsyth, K. (2004). Mechanistic
insights into the kinetics of HIV-1 nucleocapsid protein-facilitated tRN A annealing to the
primer binding site. J Mol Biol 33 7, 951-968.

Heale, B.S., Soifer, H.S., Bowers, C., and Rossi, U. (2005). siRNA target site secondary
structure predictions using local stable substructures. Nucleic Acids Res 33 , e30.

Hein, C.D., Liu, X.M., and Wang, D. (2008). Click chemistry, a powerful tool for
pharmaceutical sciences. Pharrn Res 25, 2216-2230.

Hofacker, LL. (2003). Vienna RNA secondary structure server. Nucleic Acids Res 31,
3429-3431.

Hofacker, I.L., Fontana, W., Stadler, P.F., Bonhoeffer, L.S., Tacker, M., and Schuster, P.
(1994). Fast Folding and Comparison of Rna Secondary Structures. Monatshette Fur
Chemie 125, 167-188.

Hohjoh, H. (2002). RNA interference (RNA(i)) induction with various types of synthetic
oligonucleotide duplexes in cultured human cells. FEBS Lett 521, 195-199.

Homung, V., Guenthner-Biller, M., Bourquin, C., Ablasser, A., Schlee, M., Uematsu, S.,
Noronha, A., Manoharan, M., Akira, S., de Fougerolles, A., et al. (2005). Sequence-
specific potent induction of IFN-alpha by short interfering RNA in plasmacytoid
dendritic cells through TLR7. Nat Med 11, 263-270.

Hu-Lieskovan, S., Heidel, J.D., Bartlett, D.W., Davis, M.B., and Triche, TJ. (2005).
Sequence-specific knockdown of EWS-FLIl by targeted, nonviral delivery of small

146

interfering RNA inhibits tumor growth in a murine model of metastatic Ewing's sarcoma.
Cancer Res 65, 8984-8992.

Hutvagner, G. (2005). Small RNA asymmetry in RNAi: Function in RISC assembly and
gene regulation. FEBS Letters 579, 5850-5857.

Hutvagner, G., and Sirnard, MJ. (2008). Argonaute proteins: key players in RNA
silencing. Nat Rev Mol Cell Biol 9, 22-32.

Hutvagner, G., and Zamore, PD. (2002). A microRNA in a multiple-turnover RNAi
enzyme complex. Science 297, 2056-2060.

Ishizuka, A., Siomi, M.C., and Siomi, H. (2002). A Drosophila fragile X protein interacts
with components of RNAi and ribosomal proteins. Genes Dev 16, 2497-2508.

Jackson, A.L., Bartz, S.R., Schelter, J., Kobayashi, S.V., Burchard, J ., Mao, M., Li, B.,
Cavet, G., and Linsley, RS. (2003). Expression profiling reveals off-target gene
regulation by RNAi. Nat Biotechnol 21, 63 5-637.

Jackson, A.L., Burchard, J ., Leake, D., Reynolds, A., Schelter, J ., Guo, J ., Johnson, J .M.,
Lim, L., Karpilow, J ., Nichols, K., et al. (2006). Position-specific chemical modification
of siRNAs reduces "off-target" transcript silencing. RNA 12, 1197-1205.

Jagla, B., Aulner, N., Kelly, P.D., Song, D.A., Volchuk, A., Zatorski, A., Shum, D.,
Mayer, T., De Angelis, D.A., Ouerfelli, O., et al. (2005). Sequence characteristics of
frmctional siRNAs. RNA 11, 864-872.

Jakymiw, A., Lian, S., Eystathioy, T., Li, S., Satoh, M., Hamel, J .C., Fritzler, M.J., and
Chan, E.K. (2005). Disruption of GW bodies impairs mammalian RNA interference. Nat
Cell Biol 7, 1267-1274.

Jeong, J .H., and Park, TC. (2002). Poly(L-lysine)-g-poly(D,L-lactic-co-glycolic acid)
micelles for low cytotoxic biodegradable gene delivery carriers. J Control Release 82,
159-166.

Jiang, X., Vogel, E.B., Smith, M.R., and Baker, G.L. (2008). "Clickable" Polyglycolides:
Tunable Synthons for Therrnoresponsive, Degradable Polymers. Macromolecules 41,
1937-1944.

Judge, A., and Maclachlan, I. (2008). Overcoming the innate immune response to small
interfering RNA. Hum Gene Ther 19, 111-124.

Judge, A.D., Sood, V., Shaw, J .R., Fang, D., McClintock, K., and Maclachlan, I. (2005).

Sequence-dependent stimulation of the mammalian innate immune response by synthetic
siRNA. Nat Biotechnol 23, 457-462.

147

Kariko, K., Bhuyan, P., Capodici, J ., and Weissman, D. (2004). Small Interfering RNAs
Mediate Sequence-Independent Gene Suppression and Induce Immune Activation by
Signaling through Toll-Like Receptor 3. J Irnmunol 1 72, 6545-6549.

Katoh, T., and Suzuki, T. (2007). Specific residues at every third position of siRNA
shape its efficient RNAi activity. Nucleic Acids Res, e27.

Kawahara, Y., Zinshteyn, B., Chendrimada, T.P., Shiekhattar, R., and Nishikura, K.
(2007). RNA editing of the microRNA-151 precursor blocks cleavage by the Dicer-
TRBP complex. EMBO Rep.

Khvorova, A., Reynolds, A., and J ayasena, SD. (2003). Functional siRNAs and miRNAs
exhibit strand bias. Cell 115, 209-216.

Kim, D.H., Behlke, M.A., Rose, S.D., Chang, M.S., Choi, S., and Rossi, J.J. (2005).
Synthetic dsRNA Dicer substrates enhance RNAi potency and efficacy. Nat Biotechnol
23, 222-226.

Kim, D. H. Longo, M, Han, Y., Lundberg, P. Cantin, E., and Rossi, J. J. (2004).
Interferon induction by siRNAs and ssRNAs synthesized by phage polymerase. Nat
Biotechn0122,321-325.

Kim, D.H., and Rossi, J .J . (2003). Coupling of RNAi-mediated target downregulation
with gene replacement. Antisense Nucleic Acid Drug Dev 13, 151-155.

Kim, E.J., Kramer, S.F., Hebert, C.G., Valdes, J .J., and Bentley, W.E. (2007). Metabolic
engineering of the baculovirus-expression system via inverse "shotgun" genomic analysis
and RNA interference (dsRNA) increases product yield and cell longevity. Biotechnol
Bioeng 98, 645-654.

Kim, 1., Liu, C.W., and Puglisi, J .D. (2006). Specific Recognition of HIV TAR RNA by
the dsRNA Binding Domains (dsRBDl-dsRBDZ) of PKR. J Mol Biol 358, 430-442.

Kini, H.K., and Walton, SP. (2007). In vitro binding of single-stranded RNA by human
Dicer. FEBS Lett 581, 5611-5616.

Kircheis, R., Wightrnan, L., and Wagner, E. (2001). Design and gene delivery activity of
modified polyethylenimines. Adv Drug Deliv Rev 53, 341 -358.

Kleinman, M.B., Yarnada, K, Takeda, A., Chandrasekaran, V., Nozaki, M., Baffi, J 1.,
Albuquerque, R.J., Yamasaki, S., ltaya, M., Pan, Y., et al. (2008). Sequence- and target-
independent angiogenesis suppression by siRNA via TLR3. Nature 452, 591-597.

Koberle, C., Kaufinann, SH, and Patzel, V. (2006). Selecting effective siRNAs based on
guide RNA structure. Nat Protoc 1, 1832-1839.

148

Kok, K.H., Ng, M.H., Ching, Y.P., and Jin, D.Y. (2007). Human TRBP and PACT
Directly Interact with Each Other and Associate with Dicer to Facilitate the Production of
Small Interfering RNA. J Biol Chem 282, 17649-17657.

Kolb, H., Finn, M., and Sharpless, K. (2001). Click Chemistry: Diverse Chemical
Function fiom a Few Good Reactions. Angew Chem Int Ed Engl 40, 2004-2021.

Kunath, K., von Harpe, A., Petersen, H., Fischer, D., Voigt, K., Kissel, T., and Bickel, U.
(2002). The structure of PEG-modified poly(ethylene irnines) influences biodistribution
and pharmacokinetics of their complexes with NF-kappaB decoy in mice. Pharm Res 19,
810-817.

Lai, EC. (2002). Micro RNAs are complementary to 3' UTR sequence motifs that
mediate negative post-transcriptional regulation. Nat Genet 30, 363-364.

Larnberton, J .S., and Christian, A.T. (2003). Varying the nucleic acid composition of
siRNA molecules dramatically varies the duration and degree of gene silencing. Mol
Biotechn0124, 111-120.

Laraki, G., Clerzius, G., Daher, A., Melendez-Pena, C., Daniels, S., and Gatignol, A.
(2008). Interactions between the double-stranded RNA-binding proteins TRBP and
PACT define the Medipal domain that mediates protein-protein interactions. RNA Biol 5,
92-103.

Lee, Y., Ahn, C., Han, J., Choi, H., Kim, J., Yirn, J., Lee, J., Provost, P., Radrnark, 0.,
Kim, S., et al. (2003). The nuclear RNase III Drosha initiates microRNA processing.
Nature 425, 415-419.

Lee, Y., Hur, 1., Park, S.Y., Kim, Y.K., Suh, M.R., and Kim, V.N. (2006). The role of
PACT in the RNA silencing pathway. EMBO J 25, 522-532.

Leonard, J.N., Shah, P.S., Burnett, J.C., and Schaffer, D.V. (2008). HIV Evades RNA
Interference Directed at TAR by an Indirect Compensatory Mechanism. Cell Host
Microbe 4, 484-494.

Li, S., Peters, G.A., Ding, K., Zhang, X., Qin, J ., and Sen, GO (2006). Molecular basis
for PKR activation by PACT or dsRNA. Proc Natl Acad Sci U S A 103, 10005-10010.

Lian, S., Fritzler, M.J., Katz, J., Hamazaki, T., Terada, N., Satoh, M., and Chan, E.K.
(2007). Small interfering RNA-mediated silencing induces target-dependent assembly of
GW/P bodies. Mol Biol Cell 18, 3375-3387.

Lima, W.F., Murray, H., Nichols, J.G., Wu, H., Sun, H., Prakash, T.P., Berdeja, A.R.,
Gaus, H.J., and Crooke, ST. (2009). Human Dicer Binds Short Single-strand and
Double-strand RNA with High Affinity and Interacts with Different Regions of the
Nucleic Acids. J Biol Chem 284, 2535-2548.

149

Liu, J., Carmell, M.A., Rivas, F.V., Marsden, C.G., Thomson, J.M., Song, J.J.,
Hammond, S.M., Joshua-Tor, L., and Harmon, G.J. (2004). Argonaute2 Is the Catalytic
Engine of Mammalian RNAi. Science 305, 1437-1441.

Liu, J ., Rivas, F.V., Wohlschlegel, J ., Yates, J .R., Parker, R., and Harmon, G.J. (2005a).
A role for the P-body component GW182 in microRNA function. Nat Cell Biol 7, 1261-
1266.

Liu, J ., Valencia-Sanchez, M.A., Harmon, G.J., and Parker, R. (2005b). MicroRNA-
dependent localization of targeted mRNAs to mammalian P-bodies. Nat Cell Biol 7, 719-
723.

Liu, Q., Rand, T.A., Kalidas, S., Du, F., Kim, H.E., Smith, DP, and Wang, X. (2003).
R2D2, a bridge between the initiation and effector steps of the Drosophila RNAi
pathway. Science 301, 1921-1925.

Long, D., Lee, R., Williams, P., Chan, C.Y., Ambros, V., and Ding, Y. (2007). Potent
effect of target structure on microRNA function. Nat Struct Mol Biol 14, 287.

Lu, Z.J., and Mathews, DH. (2007). Efficient siRNA selection using hybridization
thermodynamics. Nucleic Acids Res 36, 640-647.

Lund, E., Guttinger, S., Calado, A., Dahlberg, J .E., and Kutay, U. (2004). Nuclear Export
of MicroRNA Precursors. Science 303, 95-98.

Luo, K.Q., and Chang, DC. (2004). The gene-silencing efficiency of siRNA is strongly
dependent on the local structure of mRN A at the targeted region. Biochem Biophys Res
Commun 318, 303-310.

Ma, J .B., Ye, K., and Patel, DJ. (2004). Structural basis for overhang-specific small
interfering RNA recognition by the PAZ domain. Nature 429, 318-322.

Ma, J .B., Yuan, Y.R., Meister, G., Pei, Y., Tuschl, T., and Patel, DJ. (2005). Structural
basis for 5'-end-specific recognition of guide RNA by the A. fulgidus Piwi protein.
Nature 434, 666-670.

MacRae, I.J., Ma, E., Zhou, M., Robinson, C.V., and Doudna, J .A. (2008). In vitro
reconstitution of the human RISC-loading complex. Proc Natl Acad Sci U S A 105, 512-
517.

MacRae, I.J., Zhou, K., and Doudna, J .A. (2007). Structural determinants of RNA
recognition and cleavage by Dicer. Nat Struct Mol Biol 14, 934-940.

MacRae, I.J., Zhou, K., Li, F ., Repic, A., Brooks, A.N., Cande, W.Z., Adams, P.D., and

Doudna, J .A. (2006). Structural basis for double-stranded RNA processing by Dicer.
Science 311, 195-198.

150

Manche, L., Green, S.R., Schmedt, C., and Mathews, MB. (1992). Interactions between
double-stranded RNA regulators and the protein kinase DAI. Mol Cell Biol 12, 5238-
5248.

Maniataki, E., and Mourelatos, Z. (2005). A human, ATP-independent, RISC assembly
machine fueled by pre-rniRNA. Genes Dev 19, 2979-2990.

Manjunath, N., Wu, H., Subramanya, S., and Shankar, P. (2009). Lentiviral delivery of
short hairpin RNAs. Advanced Drug Delivery Reviews 61, 732-745.

Mann, A., Richa, R., and Ganguli, M. (2008). DNA condensation by poly-L-lysine at the
single molecule level: Role of DNA concentration and polymer length. J Control Release
125, 252-262.

Mao, S., Neu, M., Gerrnershaus, O., Merkel, O., Sitterberg, J ., Bakowsky, U., and Kissel,
T. (2006). Influence of Polyethylene Glycol Chain Length on the Physicochemical and
Biological Properties of Poly(ethylene imine)-<i>grafi</i>-Poly(ethylene glycol) Block
Copolymer/SiRNA Polyplexes. Bioconjug Chem 1 7, 1209-1218.

Markham, NR, and Zuker, M. (2005). DINAMelt web server for nucleic acid melting
prediction. Nucleic Acids Res 33, W577-581.

Markham, NR, and Zuker, M. (2008). UNAFold: software for nucleic acid folding and
hybridization. Methods Mol Biol 453, 3-31.

Martinez, J ., and Tuschl, T. (2004). RISC is a 5' phosphomonoester-producing RNA
endonuclease. Genes Dev 18, 975-980.

Mathews, D.H., Disney, M.D., Childs, J .L., Schroeder, S.J., Zuker, M., and Turner, DH.
(2004). Incorporating chemical modification constraints into a dynamic programming
algorithm for prediction of RNA secondary structure. Proc Natl Acad Sci U S A 101,
7287-7292.

Mathews, D.H., Sabina, J., Zuker, M., and Turner, DH. (1999). Expanded sequence
dependence of thermodynamic parameters improves prediction of RNA secondary
structure. J Mol Biol 288, 911.

Matranga, C., Tornari, Y., Shin, C., Bartel, DR, and Zamore, PD. (2005). Passenger-
Strand Cleavage Facilitates Assembly of siRNA into Ag02-Containing RNAi Enzyme
Complexes. Cell 123, 607-620.

Matveeva, O., Nechipurenko, Y., Rossi, L., Moore, 3., Saetrom, P., Ogurtsov, A.Y.,
Atkins, J .F., and Shabalina, SA. (2007). Comparison of approaches for rational siRNA
design leading to a new efficient and transparent method. Nucleic Acids Res 35, e63.

McBride, J .L., Boudreau, R.L., Harper, S.Q., Staber, P.D., Monteys, A.M., Martins, 1.,
Gilmore, B.L., Burstein, H., Peluso, R.W., Polisky, B., et al. (2008). Artificial miRNAs

151

 

mitigate shRNA-mediated toxicity in the brain: implications for the therapeutic
development of RNAi. Proc Natl Acad Sci U S A 105, 5868-5873.

Meister, G., Landthaler, M., Patkaniowska, A., Dorsett, Y., Teng, G., and Tuschl, T.
(2004). Human Argonaute2 Mediates RNA Cleavage Targeted by miRNAs and siRNAs.
Mol Cell 15, 185-197.

Meister, G., Landthaler, M., Peters, L., Chen, P.Y., Urlaub, H., Luhrmann, R., and
Tuschl, T. (2005). Identification of Novel Argonaute-Associated Proteins. Curr Biol 15,
2149-2155.

Mello, CC, and Conte, D., Jr. (2004). Revealing the world of RNA interference. Nature
431, 338-342.

Meylan, E., and Tschopp, J. (2006). Toll-like receptors and RNA helicases: two parallel
ways to trigger antiviral responses. Mol Cell 22, 561-569.

Muckstein, U., Tafer, H., Hackermuller, J., Bemhart, S.H., Stadler, PF, and Hofacker,
LL. (2006). Thermodynamics of RNA-RNA binding. Bioinfonnatics 22, 1177-1182.

Murphy, D., Dancis, B., and Brown, J .R. (2008). The evolution of core proteins involved
in microRNA biogenesis. BMC Evol Biol 8, 92.

Nallagatla, S.R., and Bevilacqua, RC. (2008). Nucleoside modifications modulate
activation of the protein kinase PKR in an RNA structure-specific manner. RNA 14,
1201-1213.

Nallagatla, S.R., Hwang, J., Toroney, R., Zheng, X., Cameron, CE, and Bevilacqua,
RC. (2007). 5'-triphosphate-dependent activation of PKR by RNAs with short stem-
loops. Science 318, 1455-1458.

Nanduri, S., Carpick, B.W., Yang, Y., Williams, BR, and Qin, J. (1998). Structure of the
double-stranded RNA-binding domain of the protein kinase PKR reveals the molecular
basis of its dsRNA-mediated activation. EMBO J 1 7, 5458-5465.

Nielsen, C.B., Shomron, N., Sandberg, R., Homstein, E., Kitzrnan, J ., and Burge, CB.
(2007). Determinants of targeting by endogenous and exogenous microRNAs and
siRNAs. RNA 13, 1894-1910.

Okarnura, K., Ishizuka, A., Siomi, H., and Siomi, MC. (2004). Distinct roles for
Argonaute proteins in small RNA-directed RNA cleavage pathways. Genes Dev 18,
1655-1666.

Overhoff, M., Alken, M., Far, R.K., Lemaitre, M., Lebleu, B., Sczakiel, G., and Robbins,

I. (2005). Local RNA target structure influences siRNA efficacy: a systematic global
analysis. J Mol Biol 348, 871-881.

152

 

Pardridge, W.M. (2007). shRNA and siRNA delivery to the brain. Adv Drug Deliv Rev
59, 141-152.

Parker, G.S., Maity, TS, and Bass, BL. (2008). dsRNA binding properties of RDE-4
and TRBP reflect their distinct roles in RNAi. J Mol Biol 384, 967-979.

Patel, R.C., Stanton, P., McMillan, N.M., Williams, BR, and Sen, CC. (1995). The
interferon-inducible double-stranded RNA-activated protein kinase self-associates in
vitro and in vivo. Proc Natl Acad Sci U S A 92, 8283-8287.

Patzel, V., Rutz, S., Dietrich, I., Koberle, C., Scheffold, A., and Kaufrnann, SH. (2005).
Design of siRN As producing unstructured guide-RNAs results in improved RNA
interference efficiency. Nat Biotechnol 23, 1440-1444.

Pellino, J .L., Jaskiewicz, L., Filipowicz, W., and Sontheimer, E.J. (2005). ATP modulates
siRNA interactions with an endogenous human Dicer complex. RNA 11, 1719-1724.

Provost, P., Dishart, D., Doucet, J ., Frendewey, D., Samuelsson, B., and Radmark, O.
(2002). Ribonuclease activity and RNA binding of recombinant human Dicer. EMBO J
21 , 5864-5874.

Pun, S.H., Tack, F., Bellocq, N.C., Cheng, J ., Grubbs, B.H., Jensen, G.S., Davis, M.B.,
Brewster, M., J anicot, M., J anssens, B., et al. (2004). Targeted delivery of RNA-cleaving
DNA enzyme (DNAzyme) to tumor tissue by transferrin-modified, cyclodextrin-based
particles. Cancer Biol Ther 3, 641-650.

Ramos, A., Grunert, S., Adams, J ., Micklem, D.R., Proctor, M.R., Freund, S., Bycroft,
M., St Johnston, D., and Varani, G. (2000). RNA recognition by a Staufen double-
stranded RNA-binding domain. EMBO J 19, 997-1009.

Rana, TM. (2007). Illuminating the silence: understanding the structure and function of
small RNAs. Nat Rev Mol Cell Biol 8, 23-36.

Rana, T.M., and Jeang, K.T. (1999). Biochemical and functional interactions between
HIV-1 Tat protein and TAR RNA. Arch Biochem Biophys 365, 175-185.

Rand, T .A., Petersen, S., Du, F., and Wang, X. (2005). Argonaute2 Cleaves the Anti-
Guide Strand of siRNA during RISC Activation. Cell 123, 621-629.

Reynolds, A., Anderson, E.M., Vermeulen, A., Fedorov, Y., Robinson, K., Leake, D.,
Karpilow, J ., Marshall, W.S., and Khvorova, A. (2006). Induction of the interferon
response by siRNA is cell type- and duplex length-dependent. RNA 12, 988-993.

Reynolds, A., Leake, D., Boese, Q., Scaringe, 8., Marshall, W.S., and Khvorova, A.
(2004). Rational siRNA design for RNA interference. Nat Biotechn0122, 326-330.

153

Rivas, F.V., Tolia, N.H., Song, J .J., Aragon, J .P., Liu, J ., Harmon, G.J., and Joshua-Tor,
L. (2005). Purified Argonaute2 and an siRNA form recombinant human RISC. Nat Struct
Mol Biol 12, 340-349.

Robb, GB, and Rana, TM. (2007). RNA Helicase A Interacts with RISC in Human
Cells and Functions in RISC Loading. Mol Cell 26, 523-537.

Rose, S.D., Kim, D.H., Amarzguioui, M., Heidel, J .D., Collingwood, M.A., Davis, M.B.,
Rossi, J .J ., and Behlke, M.A. (2005). Functional polarity is introduced by Dicer
processing of short substrate RNAs. Nucleic Acids Res 33, 4140-4156.

Rossi, J .J . (2008). Expression strategies for short hairpin RNA interference triggers. Hum
Gene Ther 19, 313-317.

Rozema, D.B., Ekena, K., Lewis, D.L., Loomis, A.G., and Wolff, J.A. (2003).
Endosomolysis by Masking of a Membrane-Active Agent (EMMA) for Cytoplasmic
Release of Macromolecules. Bioconjug Chem 14, 51-57.

Rozema, D.B., Lewis, D.L., Wakefield, D.H., Wong, S.C., Klein, J.J., Roesch, P.L.,
Bertin, S.L., Reppen, T.W., Chu, Q., Blokhin, A.V., et al. (2007). Dynamic
PolyConjugates for targeted in vivo delivery of siRNA to hepatocytes. Proc Natl Acad
Sci U S A 104, 12982-12987.

Ryter, J .M., and Schultz, SC. (1998). Molecular basis of double-stranded RNA-protein
interactions: structure of a dsRNA-binding domain complexed with dsRNA. EMBO J I 7,
7505-7513.

Saunders, LR, and Barber, G.N. (2003). The dsRNA binding protein family: critical
roles, diverse cellular functions. F aseb J 1 7, 961-983. ‘

Scacheri, P.C., Rozenblatt-Rosen, O., Caplen, N.J., Wolfsberg, T.G., Umayarn, L., Lee,
J .C., Hughes, C.M., Shanmugarn, K.S., Bhattacharjee, A., Meyerson, M., et al. (2004).
Short interfering RNAs can induce unexpected and divergent changes in the levels of
untargeted proteins in mammalian cells. PNAS 101, 1892-1897.

Schubert, S., Grunweller, A., Erdmann, V.A., and Kurreck, J. (2005). Local RNA Target
Structure Influences siRNA Efficacy: Systematic Analysis of Intentionally Designed
Binding Regions. J Mol Biol 348, 883-893.

Schwarz, D.S., Ding, H., Kennington, L., Moore, J .T., Schelter, J ., Burchard, J ., Linsley,
P.S., Aronin, N., Xu, Z., and Zamore, PD. (2006). Designing siRNA that distinguish
between genes that differ by a single nucleotide. PLoS Genet 2, e140.

Schwarz, D.S., Hutvagner, G., Du, T., Xu, Z., Aronin, N., and Zamore, PD. (2003).
Asymmetry in the assembly of the RNAi enzyme complex. Cell 115, 199-208.

154

 

Seeman, N.C., Rosenberg, J .M., and Rich, A. (1976). Sequence-specific recognition of
double helical nucleic acids by proteins. Proc Natl Acad Sci U S A 73, 804-808.

Shabalina, S.A., Spiridonov, A.N., and Ogurtsov, A.Y. (2006). Computational models
with thermodynamic and composition features improve siRN A design. BMC
Bioinformatics 7, 65.

Shah, J .K., Garner, H.R., White, M.A., Shames, D.S., and Minna, J .D. (2007). SIR:
siRN A Information Resource, a web-based tool for siRN A sequence design and analysis
and an open access siRNA database. BMC Bioinform 8, 178.

Shao, Y., Chan, C.Y., Maliyekkel, A., Lawrence, C.E., Roninson, LB, and Ding, Y.
(2007). Effect of target secondary structure on RNAi efficiency. RNA 13, 1631-1640.

Sharpless, KB, and Kolb, HQ (1999). Click chemistry. A concept for merging process
and discovery chemistry. Abstracts of Papers of the American Chemical Society 217,
U95-U95.

Siolas, D., Lerner, C., Burchard, J ., Ge, W., Linsley, P.S., Paddison, P.J., Harmon, G.J.,
and Cleary, M.A. (2005). Synthetic shRNAs as potent RNAi triggers. Nat Biotechnol 23,
227-231.

Siomi, H., and Siomi, M.C. (2009). On the road to reading the RNA-interference code.
Nature 45 7, 396-404.

Sioud, M. (2006). Innate sensing of self and non-self RNAs by Toll-like receptors.
Trends Mol Med 12, 167-176.

Sipa, K., Sochacka, E., Kazrnierczak-Baranska, J ., Maszewska, M., J anicka, M., Nowak,
G., and Nawrot, B. (2007). Effect of base modifications on structure, thermodynamic
stability, and gene silencing activity of short interfering RNA. RNA 13, 1301-1316.

Sledz, C.A., Holko, M., de Veer, M.J., Silverrnan, RH, and Williams, BR. (2003).
Activation of the interferon system by short-interfering RN As. Nat Cell Biol 5, 834-839.

Soifer, H.S., Sano, M., Sakurai, K., Chomchan, P., Saetrom, P., Sherman, M.A.,
Collingwood, M.A., Behlke, M.A., and Rossi, J .J. (2008). A role for the Dicer helicase
domain in the processing of thermodynamically unstable hairpin RNAs. Nucleic Acids
Res 36, 6511-6522.

Song, J .J., Smith, S.K., Harmon, G.J., and Joshua-Tor, L. (2004). Crystal Structure of
Argonaute and Its lrnplications for RISC Slicer Activity. Science 305, 1434-1437.

Sontheimer, E.J. (1994). Site-specific RNA crosslinking with 4-thiouridine. Mol Biol
Rep 20, 35-44.

155

Stein, CA. (1998). How to design an antisense oligodeoxynucleotide experiment: a
consensus approach. Antisense Nucleic Acid Drug Dev 8, 129-132.

Swami, A., Kurupati, R.K., Pathak, A., Singh, Y., Kumar, P., and Gupta, KC. (2007). A
unique and highly efficient non-viral DNA/siRNA delivery system based on PEI-
bisepoxide nanoparticles. Biochem Biophys Res Commun 362, 835-841.

Tahbaz, N., Kolb, F.A., Zhang, H., Jaronczyk, K., Filipowicz, W., and Hobman, TC.
(2004). Characterization of the interactions between mammalian PAZ PIW I domain
proteins and Dicer. EMBO Rep 5, 189-194.

Takashirna, Y., Saito, R., Nakajirna, A., Oda, M., Kimura, A., Kanazawa, T., and Okada,
H. (2007). Spray-drying preparation of microparticles containing cationic PLGA
nanospheres as gene carriers for avoiding aggregation of nanospheres. Int J Pharm 343,
262-269.

Tang, G.L., Galili, G., and Zhuang, X. (2007). RNAi and microRNA: breakthrough
technologies for the improvement of plant nutritional value and metabolic engineering.
Metabolorrrics 3, 357-369.

Terrazas, M., and K001, ET. (2009). RNA major groove modifications improve siRNA
stability and biological activity. Nucleic Acids Res 37, 346-353.

Tornari, Y., Matranga, C., Haley, B., Martinez, N., and Zamore, PD. (2004). A protein
sensor for siRNA asymmetry. Science 306, 1377-1380.

Tseng, Y.C., Mozurndar, S., and Huang, L. (2009). Lipid-based systemic delivery of
siRNA. Adv Drug Deliv Rev 61 , 721-731.

Ui-Tei, K., Naito, Y., Takahashi, F., Haraguchi, T., Ohki-Hamazaki, H., Juni, A., Ueda,
R., and Saigo, K. (2004). Guidelines for the selection of highly effective siRNA
sequences for mammalian and chick RNA interference. Nucleic Acids Res 32, 936-948.

Ui-Tei, K., Naito, Y., Zenno, S., Nishi, K., Yamato, K, Takahashi, F., Juni, A., and
Saigo, K. (2008). Functional dissection of siRNA sequence by systematic DNA
substitution: modified siRNA with a DNA seed arm is a powerful tool for mammalian
gene silencing with significantly reduced off-target effect. Nucleic Acids Res 36, 2136-
2151.

Urban-Klein, B., Werth, S., Abuharbeid, S., Czubayko, F., and Aigner, A. (2005). RNAi-
mediated gene-targeting through systemic application of polyethylenimine (PEI)-
complexed siRNA in vivo. Gene Ther 12, 461-466.

van Rij, RP, and Berezikov, E. (2009). Small RNAs and the control of transposons and
viruses in Drosophila. Trends Microbiol 1 7, 163-171.

156

 

Venter, J.C., Adams, M.D., Myers, B.W., Li, P.W., Mural, R.J., Sutton, G.G., Smith,
H.O., Yandell, M., Evans, C.A., Holt, R.A., et al. (2001). The sequence of the human
genome. Science 291, 1304-1351.

Vert, J .P., Foveau, N., Lajaunie, C., and Vandenbrouck, Y. (2006). An accurate and
interpretable model for siRN A efficacy prediction. BMC Bioinform 7, 520.

Vickers, T.A., Koo, S., Bennett, C.F., Crooke, S.T., Dean, N.M., and Baker, BR (2003).
Efficient reduction of target RNAs by small interfering RNA and RNase H-dependent
antisense agents. A comparative analysis. J Biol Chem 278, 7108-7118.

Vickers, T.A., Lima, W.F., Nichols, J .G., and Crooke, ST. (2007). Reduced levels of
Ag02 expression result in increased siRNA competition in mammalian cells. Nucleic
Acids Res 35, 6598-6610.

Vickers, T.A., Wyatt, J .R., and Freier, SM. (2000). Effects of RNA secondary structure
on cellular antisense activity. Nucleic Acids Res 28, 1340-1347.

Wakefield, D.H., Klein, J .J ., Wolff, J .A., and Rozema, DB. (2005). Membrane activity
and transfection ability of amphipathic polycations as a function of alkyl group size.
Bioconjug Chem 16, 1204-1208.

Walton, S.P., Stephanopoulos, G.N., Yarmush, ML, and Roth, CM. (1999). Prediction
of antisense oligonucleotide binding affinity to a structured RNA target. Biotechnol
Bioeng 65, 1-9.

Walton, S.P., Stephanopoulos, G.N., Yarmush, ML, and Roth, CM. (2002).
Thermodynamic and kinetic characterization of antisense oligodeoxynucleotide binding
to a structured mRNA. Biophys J 82, 366-377.

Wang, X., Xu, W., Mohapatra, S., Kong, X., Li, X., Lockey, RF, and Mohapatra, SS.
(2008). Prevention of airway inflammation with topical cream containing irrriquimod and
small interfering RNA for natriuretic peptide receptor. Genet Vaccines Ther 6, 7.

Watanabe, T., Totoki, Y., Toyoda, A., Kaneda, M., Kuramochi-Miyagawa, S., Obata, Y.,
Chiba, H., Kohara, Y., Kono, T., Nakano, T., et al. (2008). Endogenous siRNAs fiom
naturally formed dsRNAs regulate transcripts in mouse oocytes. Nature 453, 539-543.
Weiss, J .N. (1997). The Hill equation revisited: uses and misuses, pp. 835-841.

Weitzer, S., and Martinez, J. (2007). The human RNA kinase hClpl is active on 3'
transfer RNA exons and short interfering RNAs. Nature 44 7, 222-226.

Westerhout, E.M., and Berkhout, B. (2007). A systematic analysis of the effect of target
RNA structure on RNA interference. Nucleic Acids Res 35, 4322-4330.

157

 

Westerhout, E.M., Ooms, M., Vink, M., Das, A.T., and Berkhout, B. (2005). HIV -1 can
escape from RNA interference by evolving an alternative structure in its RNA genome.
Nucleic Acids Res 33, 796-804.

Wolff, J .A., and Rozema, DB. (2008). Breaking the Bonds: Non-viral Vectors Become
Chemically Dynamic. Mol Ther 16, 8-15.

Wurm, RM. (2004). Production of recombinant protein therapeutics in cultivated
mammalian cells. Nat Biotechnol 22, 1393-1398.

Xiong, M.P., Bae, Y., Fukushima, S., Forrest, M.L., Nishiyarna, N., Kataoka, K., and
Kwon, GS. (2007). pH-responsive Multi-PEGylated dual cationic nanoparticles enable
charge modulations for safe gene delivery. ChemMedChern 2, 1321-1327.

Yezhelyev, M.V., Qi, L., O'Regan, R.M., Nie, S., and Gao, X. (2008). Proton-sponge
coated quantum dots for siRNA delivery and intracellular imaging. J Am Chem Soc 130,
9006-9012.

Yoneyama, M., Kikuchi, M., Natsukawa, T., Shinobu, N., Irnaizumi, T., Miyagishi, M.,
Taira, K., Akira, S., and Fujita, T. (2004). The RNA helicase RIG-I has an essential
fimction in double-stranded RNA-induced innate antiviral responses. Nat Irnmunol 5,
730-737.

Yoshinari, K., Miyagishi, M., and Taira, K. (2004). Effects on RNAi of the tight
structure, sequence and position of the targeted region. Nucleic Acids Res 32, 691-699.

Zhang, S., Zhao, B., Jiang, H., Wang, 8., and Ma, B. (2007). Cationic lipids and
polymers mediated vectors for delivery of siRNA. J Control Release 123, 1-10.

Zheng, X., and Bevilacqua, RC. (2004). Activation of the protein kinase PKR by short
double-stranded RNAs with single-stranded tails. RNA 10, 1934-1945.

Zhu, X., and Galili, G. (2004). Lysine metabolism is concurrently regulated by synthesis
and catabolism in both reproductive and vegetative tissues. Plant Physiol 135, 129-136.

Zuhom, I.S., Engberts, J .B., and Hoekstra, D. (2007). Gene delivery by cationic lipid
vectors: overcoming cellular barriers. Eur Biophys J 36, 349-362.

Zuker, M. (2003). Mfold web server for nucleic acid folding and hybridization prediction.
Nucleic Acids Res 31, 3406-3415.

158